Introduction

The supralittoral sandhopper Talitrus saltator (Montagu 1808) evolved nocturnal habits to avoid high temperature and low humidity conditions which can rapidly lead to the animal’s death. During the day it lives buried in the wet sand of the beach. However, stressing factors such as predators or changes in the sea level can cause active displacements of the sandhopper from its preferred zone along the sea-land axis of the beach perpendicular to the shoreline (characteristic for each population dependently on the orientation of the shore (Papi and Pardi 1953). It is also well known that the sandhopper orientates seaward when exposed to dry conditions whereas submersion motivates it to direct landward (Pardi and Papi 1952).

In its zonal recovery, T. saltator is known to rely on several visual cues such as the sun and the moon (Papi and Pardi 1953; Pardi and Papi 1952, Papi and Pardi 1953) and the landscape (Williamson 1951; Ugolini et al. 1986, 2006; Edwards and Naylor 1987). Evidence for time-compensated solar and lunar orientation has been provided since the 1950s (see Pardi and Ercolini 1986 for a review). Recently, the skylight intensity gradient has been found to provide reliable directional information to this species (Ugolini et al. 2009). Furthermore, the solar orientation of T. saltator depends on a skylight cue perceived in the UV-blue band whose nature has not been identified yet. Indeed, when the perception of wavelengths shorter than 450 nm is prevented the sandhopper loses its capability to recognize the sun as a compass cue and exhibits positive phototactic tendencies (Ugolini et al. 1993, 1996). The sensitivity of the sandhopper to UV-blue light has recently clearly confirmed by electroretinogram recordings combined with selective light adaptation techniques (Ugolini et al. 2010).

Although previous studies on the spectral sensitivity of amphipods reported a single-pigment visual system (Pontoporeia affinis and P. femorata; Donner 1971, Abyssorchomete prebs; Cohen and Frank 2006, Scina crassicornis; Cohen and Frank 2007), the studies on Talitridae e suggest the existence of multiple visual pigments. In fact, discrete photoreceptors exist in the compound eye of T. saltator sensitive to UV-blue (λ = 390–450 nm) and green (λ = 500–550 nm) wavelengths, respectively. Our previous studies also showed the capability of the sandhopper to perceive polarized light in the blue band (λ = 435 nm) using a binary choice experiment while it does not use the skylight polarization pattern as a compass cue (Ugolini et al. 2013). However, the role of this factor in the celestial orientation of T. saltator has not been clarified. In Talorchestia longicornis it has been demonstrated that orientation during zonal recovery and activity rhythm entrainment are wavelength-specific behaviours. In particular wavelengths of 420–430 nm and about 520 nm are used for celestial sun compass orientation and to entrain the circadian activity rhythm, respectively (Forward et al. 2009a, b; Cohen et al. 2010). A role in the entrainment of activity rhythm could also be played by the intracerebral ocelli found in T. saltator (Frelon-Raimond et al. 2002) but this is only a mere speculation since we know nothing about spectral sensitivity and neuronal connections of this intracerebral structure. Moreover, T. longicornis perceives linearly polarized light for wavelength around 436 nm. More or less this is also true for Platorchestia platensis (Cohen and Putts 2013). Both these species are phylogenetically not too far from T. saltator, share similar problems of surviving in supralittoral environments, and similar spectral sensitivities. Therefore, it seems that a visual system based on multiple pigments is common to talitrid amphipods. However, no investigation has been made to test the presence of a regional specialization in the compound eyes of talitrid amphipods.

Considering the existence of the sensitivity in the UV-blue region in T. saltator, it might be natural to think that this may be related to the research on insects which possess UV-blue photoreceptors (Cronin and Bok 2016) and represent a remarkable example of a highly efficient spatial orientation capacity relying on celestial visual cues such as solar and lunar positions and the skylight polarization pattern. In contrast with T. saltator, the existence of a specialized dorsal area in insects compound eyes (dorsal rim area, DRA) is well known, and several investigations demonstrated the involvement of ommatidia occurring in this region in the detection of the polarized light (see Horváth and Varjù 2004; Wehner and Labhart 2006 for reviews). Moreover, short-wavelength receptor cells are mainly located in the DRA to optimize the reliability of the celestial polarization pattern for their orientation behaviour, whereas those perceiving longer wavelengths occur in the rest of the eye (Horváth and Varjù 2004; Cronin et al. 2014). Herrling (1976) reported ultrastructural regionalization in the compound eye of the desert ant Cataglyphis bicolor with three different retinular types, and discussed the type-I ommatidia which exist mainly in the ventral part of the compound eye used to detect far horizon landmarks for its navigation. Those reports showed that insects possess functional different regions in their compound eye serving specific behavioural roles.

The aim of this work is to assess the eventual regionalization of the visual capabilities involved in celestial compass orientation in the sandhopper T. saltator through morphological, electrophysiological and behavioural investigations.

Methods

Adult individuals of T. saltator were collected on a sandy beach in the Regional Natural Park of Migliarino, San Rossore, Massaciuccoli, Pisa, Italy (43°40′03′′ N, 10°20′29′′ E, sea-land axis = 265°–85°) over spring–summer 2016. After the collection, animals were transported to the laboratory of the University of Florence and kept in Plexiglas boxes containing damp sand; food (universal dried fish food, SERA, Vipan, Heisemberg, Germany) was available ad libitum. Sandhoppers were maintained at ambient temperature (25 °C) under an L:D = 12:12 cycle in phase with the natural photoperiod (sunrise = 0600 h, sunset = 1800 h) and the behavioural experiments were performed within 14 days from their capture. All the tests were carried out with intact individuals and with individuals with the dorsal (1/3) or ventral (2/3) part of their eyes painted by black enamel (Fig. 1c, d). For the morphological and electrophysiological experiments, animals were transferred by airplane to Japan and examined at Hamamatsu University School of Medicine.

Fig. 1
figure 1

Schematic representations of the holding method for electrophysiological recording. a The animal cut the second abdominal segment was introduced into the plastic tube and the head was glued by beeswax and resin to separate electrically. b A part of the compound eye was pained and the tip of the electrode was placed just below the cornea at the centre of an unpainted area. (PA) preamplifier. c, d Sandhoppers with the dorsal 1/3 region or 2/3 ventral region of their eyes painted, respectively. e Transmittance curve of the Blue gelatine filter

Full size image

Morphological investigations by microscopy

Scanning Electron Microscope (SEM)

The NanoSuit method was used to observe T. saltator (Ohta et al. 2014). Briefly, the whole body of living sandhoppers was dipped into the commercially available NanoSuit solution (Nissin EM Co., Ltd) for ca.10 s, blotted briefly on a dry filter paper to remove excess solution. Then, the living specimen was introduced into the SEM to construct a NanoSuit membrane, a very thin film on the surface of the specimen, by the irradiation of electron beams, without performing any traditional treatments such as chemical fixation or dehydration. The head of the living sandhopper was observed by FE-SEM (Hitachi S-4800) at an acceleration voltage of 5.0 kV.

Transmission electron microscope (TEM)

The fine structure of the compound eye has been observed in day condition,. The compound eye is easily broken by the pressure from scissors or a scalpel when attempting to remove the eye from the body. Therefore, the head was removed gently from the body in primary fixative solution (2% glutaraldehyde with 2% paraformaldehyde in 0.1 mol/l sodium cacodylate buffer, pH 7.2) with a small piece of razor blade, and kept for 2 h at 4 °C. The fixed heads were dissected through the median line of the head in a buffered solution, and were then postfixed for 1 h in 1% OsO4 in 0.1 mol/l cacodylate buffer and washed three times for 10 min each in 0.1 mol/l cacodylate buffer. Dehydration through a graded series of ethanol solutions was followed by embedding in an Epon–Araldite mixture. Ultrathin (approximately 70 nm) sections were cut perpendicular or parallel to the optical axis of each ommatidium and stained with 2% uranyl acetate followed by 0.4% lead citrate for 5 min each. The position of the cross-sections compared at different positions of the compound eye was 10 μm below the proximal edge of the crystalline cones layer. Observations were performed with JEOL JEM-1400 transmission electron microscope (100 kV).

Measurement of the visual field of the unpainted area

To measure the vertical visual filed of the unpainted area both of the DP (painted the 2/3 ventral part of the compound eye, Fig. 1d) and VP (painted the 1/3 dorsal part, Fig. 1c), we made light microscopic observations of the semi-thin section of the eye at the centre of the compound eye in the transverse plane. The embedded specimen in an Epon–Araldite mixture same as the TEM observation were sectioned parallel to the optical axis of each ommatidium and stained with Toluidine blue. We lined the optical axis between the centre of the proximal region and the distal end of the crystalline cone at both ends of the unpainted area (Fig. 3a, d) to find the line of the dichotomy between the dorsal and ventral visual field predicted.

Electrophysiological recordings

Because of the easy deformation of the eye structure when the body is pressed, the animal was gently treated not to give any mechanical stress to the eyes. Therefore, the whole animal was used but we cut it from the 2nd abdominal segment in the seawater. The cut body was put into the plastic tube, and placed only the head part including the compound eyes out of the tube. The bodyside of the head was glued to a tube using melted beeswax and resin (1:1) (Fig. 1a). The body was placed inside the tube, and the tube was filled with artificial seawater. The indifferent electrode, a chloride silver wire, was placed into the tube (Fig. 1b) to connect electrically through the cut area of the body.

To compare the spectral response curve obtained from the dorsal part (DP) and the ventral part (VP), we prepared individuals with dorsal or ventral parts of the eyes painted black with enamel (Touch up paint T-13, SOFT99 Corp.) the dorsal or ventral parts of the eyes. To distinguish the response according to the region of DP and VP, the recording electrode was placed around the centre of the unpainted area. A small hole (ca. 50 µm, crystalline cone diameter) was made on the surface of the cornea using a razor blade, and the tip of the glass electrode filled with artificial seawater was inserted into the hole just on the surface of the crystalline cone layer for electroretinogram (ERG) recordings. The end of the light guide was placed near the compound eye of the unpainted area. With this location of electrodes, we obtained stable cornea negative responses. We also excluded the responses of on- and off-transients. Because transient responses are the pooled activity of second order neurons of the lamina, whereas the sustained photoreceptor response is the pooled activity of photoreceptors and is easily modified by transient responses (Heisenberg 1971).

Responses were amplified with a high-impedance preamplifier (Nihon Kohden MZ 8201) and a high-gain amplifier (Nihon Kohden AVH-10). The magnitude of the responses (peak amplitude) was monitored on an oscilloscope (Nihon Kohden VC10). Permanent recordings were made with a data logger (Keyence Corp. NR-600). An optical system was used, with a 500-W Xenon arc lamp (Ushio Inc., Type UXL-500D-O, Japan) with a regulated power supply (Sanso XD-25, Japan). Quartz lenses produced a parallel beam of light which passed through a heat-absorption filter (Toshiba IRA-25S), and a set of 17 narrow-band interference colour filters (Vacuum Optics Corp., Japan) scanning a range of 330–650 nm having half-bandwidth maxima of 15 nm. The light intensities of the testing were measured with a silicon photodiode (S876-1010BQ, Hamamatsu Photonics K.K.) calibrated by Hamamatsu Photonics K.K. using a photo-electron bulb, and each monochromatic light was adjusted with the aid of several neutral density filters so that each contained an equal number of photons. At all wavelengths, the maximum irradiances available at the surface of the eye were 1.3 × 1014 quanta/cm2/s. The light intensity of the white light was altered with quartz neutral-density filters (Vacuum Optics Corp., Japan) to obtain intensity-response (V-log I) curves. To get the spectral-response curves, the light irradiance of each testing light was 1.3 × 1013 quanta/cm2/s, which corresponds with 1 log unit lower than the maximum irradiance. Each beam was interrupted by shutters (Uniblitz photographic and Copal) controlled for the duration of the test flash and adaptation times by a stimulator (Nihon Kohden SEN-7103). The eye was stimulated by a 200-ms pulse of light, and the stimulus interval was 20 s. Experiments were performed from 12:00 to 17:00 h during the day. All the preparations for electrophysiology were performed under light to keep the day condition; therefore, the dark periods were limited to the only light stimulus intervals for the experiments.

Data for the spectral sensitivity were plotted as the reciprocal of the irradiance required to evoke the criterion response at each wavelength, and normalized to the wavelength of maximum sensitivity at 450 nm in the spectral response curve obtained from the ventral part. Templates for A1-based visual pigment (rhodopsin) absorbance (Govardovskii et al. 2000) were used to calculate the absorptance curve. To compare the spectral sensitivity curve obtained by ERG method and the visual pigment absorptance curve, we inserted the absorptance curves calculated at 390, 430, 450 and 550 nm and normalize them to the peak sensitivities of each curve.

Celestial compass experiments

Behavioural experiments were carried out in a confined environment using a device (Fig. 1e) similar to that described by Ugolini and Macchi (1988). It consisted on a tripod supporting a horizontal transparent plate (diameter = 30 cm) with a Plexiglas bowl (diameter = 20 cm) placed on where sandhoppers were released. A goniometer was set under the bowl to detect the directions assumed by individuals. The vision of the surrounding landscape was prevented using an opaline Plexiglas screen (diameter = 30 cm, height = 3 cm).

Sandhoppers were dehydrated for a few minutes before being tested to motivate them to orientate seaward (i.e. towards 265°). Groups of 10-15 individuals were released into the bowl at a time since it has been demonstrated their non-reciprocal influence in performing directional choices (Scapini et al. 1981). It is well known that sandhoppers released in a confined environment under the celestial references jump and/or walk slipping on the perimeter of the bowl with their heads facing the outside. This allows us to make a distinction between radial (with the head pointed toward the outside of the bowl and the longitudinal axis of the body-oriented no more than ± 45° from the radius of the bowl) or non-radial individuals (they turn around the perimeter of the bowl).

A single direction (error ± 2.5°) for each radially-oriented animal was measured from freeze-frame images recorded using a video-camera placed under the apparatus. In each trial, even the number of radially orientated individuals out of the total number of individuals released was registered.

To assess the eventual regionalization in the perception of celestial cues within the compound eye of this species and their use in zonal orientation, we tested separately both intact individuals and sandhoppers subjected to the occlusion of discrete regions of their eyes. In particular, we performed the following experimental treatments:

  1. 1.

    painting of the dorsal part (DP, corresponding to the 1/3 upper area) of their eyes with black enamel (Rainbow, Maimeri, S.p.A., Mediglia, Milano, Italy) (Fig. 1c).

  2. 2.

    painting of their whole eyes except for the dorsal edge (the 2/3 ventral part, VP) (Fig. 1d).

Releases were carried out both under the natural sky and with the superimposition of a gelatine blue filter (Moonlight Blue, no 183, Spotlight, Milano, Italy, λmax = 450 nm, transmittance = 73%, Fig. 1e). We decided to use the blue filter because it is known that for wavelengths around 450 nm T. saltator perceives some celestial orienting factors (Ugolini et al. 1993, 1996, 2009, 2010, 2013).

Tests were repeated in conditions of sun visible and with its vision prevented using a screen (42 × 42 cm) placed at a distance of about 2 m from the bowl.

Data were analysed according to the methods proposed by Batschelet (1981) for circular distributions.

The length of the mean resultant vector (r) and the mean angle (α) were calculated. To establish whether the distributions differed statistically from uniformity the V test was used (p < 0.05 at least).

Furthermore, to evaluate if discrete distributions were statistically different from each other we carried out pairwise comparisons using Watson’s U2 test (p < 0.05 at least).

Since the frequency of radially-orientated individuals is considered a good indicator to assess the difficulty of sandhoppers in their directional choices we conducted statistical comparisons between frequencies recorded in different trials using the G test (p < 0.05 at least) (Zar 1984).

Results

Morphological investigations by microscopy

Optical and scanning electron microscopy (SEM)

The compound eye of T. saltator is sessile and under the light microscope each ommatidium shows up as a black dot (Fig. 2a; see Mayrat 1981). The cornea of each ommatidium as well as that of the compound eye is flat. The surface of the cornea covering the eye is smooth and not divided into facets as in many other compound eyes when observed by the NanoSuit method using a SEM (Fig. 2b).

Fig. 2
figure 2

Anatomical and ultrastructural organization of the eye. a A light microscopic view of the head shows each ommatidium through the transparent cornea. b Scanning electron microscope of the region of compound eye observed by the NanoSuit method shows the smooth surface of the cornea and not divided into facets as in many other compound eyes. c Transmission electron micrograph of the transverse section of an ommatidium. The ommatidium is a fused-type rhabdom consisting of five photoreceptors (R1–R5). d–i Transverse sections around 10 μm below the proximal end of the crystalline cone; the schematic drawings (d, f, h) and the transmission electron micrographs of the rhabdom (e, g, i), obtained from the dorsal region (d, e), median region (around the equatorial line) (f, g), and the ventral region (h, i) of the compound eye. White arrows indicate each rhabdomere of R5; black arrows indicate the soma of R5. j transmission electron microscope of the longitudinal section of ommatidia. k Enlarged image of microvilli. The scale bars indicate 500 μm in (a, b), 50 μm in (c), 1 μm in (e, g, i), 50 μm in (j), and 1 μm in (k). Cr cornea, Cc crystalline cone, Rh rhabdom, Re retinular cell, Bm basement membrane

Full size image

Transmission electron microscopy (TEM)

TEM observation of the longitudinal section revealed that each ommatidium is located beneath an undifferentiated cuticular lens and the existence of the long crystalline cone. The length of the crystalline cone occupied half of the ommatidium (Fig. 2j). Following the crystalline cone, there are rhabdoms produced by retinular cells without a crystalline tract, therefore the compound eye of T. saltator is classified as apposition-type.

Between the proximal end of the crystalline cone and the depth of 5 μm, the rhabdom consists of four retinular cells (R1–4) (data not shown). The smaller R5 retinular cells were observed from around 5 μm below the proximal end of crystalline cone; therefore, we compared the cross sections of the rhabdom 10 μm below the crystalline cone layer. They show the radially arranged five retinular cells (R1–R5) producing a square rhabdom, in the centre of the retinular cells in each ommatidium. It is obvious that the square rhabdom is produced by four triangular rhabdomeres of each retinular cell (R1–4). The sizes of them are almost equal, and the microvillar orientation of the pair R1 and R3 is orthogonal to the pair R2 and R4 (Fig. 2c, schema d, f, h). The smaller R5 exists between R1 and R4, and the microvillar orientation of the rhabdomere is parallel to the dorsoventral line of the compound eye. The area ratio of the R5 rhabdomere (Fig. 2e, g, i, white arrows) to the whole rhabdomeres (R1–4 and R5) was 6.90 ± 0.14% (n = 5) in the dorsal part (Fig. 2d, e), and 1.94 ± 0.41% (n = 5) in the median part (Fig. 2f, g). In the ventral part (Fig. 2h, i), the areal ratio was decreased to 0.40 ± 0.28% (n = 5). The enlarged image of the microvilli shows that the orientation was perpendicular to the optical axis (Fig. 2k). The microvillar orientation of R1 and R3 are parallel, and that of R4 and R6 are also parallel, and this parallel microvillar orientation is orthogonal between the two pairs (Fig. 2c, d, f, h).

The schematic drawings of Fig. 3a–e show the visual fields of DP (Fig. 3a–c) or VP (Fig. 3d, e) when they were covered by black paint. Each Θ of DP was 55°, therefore the animal might be able to obtain a visual field of ca. 90° using both eyes. In contrast, the visual field of each VP was ca. 90°. Therefore, we painted the eye of 1/3 dorsal part (DP) or 2/3 ventral part (VP) by black paint in electrophysiological and behavioural experiments.

Fig. 3
figure 3

Visual field in dorsal and ventral regions. a, d light microscopy view of the longitudinal sections of the compound eye along the transverse plane. Scale bars indicate 100 μm. b, c unpainted dorsal part. e unpainted ventral part. Small crystalline cones were omitted. Dorsal part Θ=55°; α=90°; Ventral part; Θ′=90°

Full size image

Electrophysiological recordings

Compound eyes with DP or VP regions painted black showed corneal-negative sustained waveform when subjected to around 1.0 × 1013 quanta/cm2/s. However, on-transient waveforms were observed above the light intensity at the wavelength of peak sensitivities such as UV-blue or green region. Therefore, we stimulated with light intensities, which produced sustained waveforms (Fig. 4b, insert).

Fig. 4
figure 4

Electrophysiological recordings. a The V/log I function obtained for the dorsal (black dots) and ventral (white dot) part of the compound eye plotted from responses to 200 ms flashes of varying intensity of white light. Data are means (± SD, n = 5). b Normalized, averaged (± SD, n = 5) spectral sensitivity curves from dorsal (black dots) and ventral (white dots). Representative ERG traces are shown in the insert with the positive square wave indicating stimulus duration. Dotted lines represent absorbance spectra of visual pigments that contain retinal (A1) as a chromophore (Govardovskii et al. 2000) with corresponding λmax in 390, 430, 450 and 550 nm which correspond to the peaks of ERG spectral sensitivity curves. The insert shows an example of the sustained waveforms

Full size image

The relative response (V/Vmax) curves (V-log I curves) and the relative spectral sensitivity curves derived from ERG recordings are shown in Fig. 4. There was no apparent difference at the initial slope, but a slight difference at the saturation level between dorsal and ventral parts in V-log I curves (Fig. 4a). In the DP of the compound eye, the spectral sensitivity curves showed two distinct maxima, one in the long-wave range peaking at 550 nm and one in the UV-blue range peaking at 430 nm, divided by a sensitivity minimum at approximately 470 nm (Fig. 4b). The spectral sensitivities of VP differed significantly from those of DP. The spectral sensitivity curves of VP possessed a highly sensitive region only around UV-blue peaking at 390 and 450 nm, decreasing gradually at longer wavelengths; no apparent peak was observed. The absorbance spectra of visual pigments that contain retinal as a chromophore (Govardovskii et al. 2000) corresponding with λmax in 390, 430, 450 and 550 nm are shown in Fig. 4b. These curves are all narrower than the spectral sensitivity curves of DP and VP.

Celestial compass experiments

In the releases conducted under the natural sun and sky not treated individuals (controls) exhibited a mean orientation in good agreement with the seaward direction of their home beach (Fig. 5a). Similarly, sandhoppers with either the dorsal part (DP) (Fig. 5b) or the 2/3 VP (Fig. 5c) of their eyes obscured performed directional choices mainly directed toward the expected direction. Pairwise comparisons between these distributions using Watson’s U2 test did not reach the statistical significance in any case. Instead, the difference in the frequencies of radially-orientated individuals is statistically significant (G = 9.745, df = 2, p < 0.01, G test, Fig. 6a).

Fig. 5
figure 5

Releases under the natural sun and sky. a, d, g, j Intact individuals. b, e, h, k Individuals with the 1/3 dorsal region of their eyes painted. c, f, i, l Individuals with the 2/3 ventral region of their eyes painted. N North; black arrows, mean vectors and angles (the length of the mean vector ranges from 0 to 1 = radius of the circle); black dots, sandhoppers’ directions (each dot corresponds to the direction of one individual); black triangles outside the distributions, seaward direction of animals’ home beach. The symbol of the sun indicates the solar azimuth at the time of releases. n/nt, number of radially-orientated sandhoppers (n) out of the total individuals tested (nt). The values of the V test, u, and the probability level, p, are also given

Full size image
Fig. 6
figure 6

Frequencies of radially orientated sandhoppers. a Natural sun and sky. b Natural sky with the vision of the sun prevented. c Blue filter with the vision of the sun allowed. d Blue filter with the vision of the sun prevented. The total number of individuals released is given at the top of each bar. See text for further explanations

Full size image

When tested under the natural sky and with the sun screened, controls individuals (Fig. 5d) and sandhoppers with the VP of their eyes painted (Fig. 5f) orientated in accordance with the seaward direction of their home beach (14° and 0° of difference, respectively). Even animals with the DP exhibited a mean orientation in accordance with the expected direction (Fig. 5e) but the difference between the seaward direction and the mean resultant angle is 28° and they were more dispersed than the other groups of individuals. Comparisons between distributions revealed that the distribution obtained from releases of sandhoppers with the DP of their eyes painted differed statistically from those referring to experiments conducted with intact animals (Fig. 5e vs d, U2(26, 35)  = 0.201, p < 0.05, Watson’s U2 test) and individuals with the 2/3 ventral area of their eyes obscured (Fig. 5e vs f, U2(26, 9)  = 0.205, p < 0.05 Watson’s U2 test). In this case, the difference in the frequencies of radially orientated sandhoppers did not reach the statistical significance (G = 2.778, df = 2, p = n.s., Fig. 6b, G test).

In experiments carried out under the blue gelatine filter and the vision of the sun allowed, controls exhibited a mean orientation significantly concordant with the expected direction even though deflected by 25° (Fig. 5g). Even the directional choices performed by individuals subjected to the occlusion of either the DP (Fig. 5h) or the VP (Fig. 5i) of their eyes were mainly orientated in accordance with the seaward direction of their home beach (28° and 24° of difference, respectively). Pairwise comparisons carried out by the Watson’s U2 test revealed that the distributions did not differ statistically each other. Instead, significant differences in the frequencies of radially-orientated sandhoppers were pointed out (G = 6.784, df = 2, p < 0.05, G test, Fig. 6c).

In tests of spectral filtering carried out with the vision of the sun prevented, intact sandhoppers (Fig. 5j) and animals with the VP of their eyes occluded (Fig. 5l) were mainly orientated in accordance with the expected direction (7° and 34°, respectively). Instead, the mean orientation of individuals with the DP of their eyes black-painted (Fig. 5k) was not concordant with their escape direction. Indeed, although the distribution obtained from these releases reaches the statistical significance, the mean resultant vector was deflected by 52° from the expected direction and most sandhoppers subjected to this treatment are unable to head in a direction whatever: they turn around the perimeter of the bowl (49%, n = 67, Fig. 5k and Fig. 6d). Moreover, comparisons pointed out that the distribution referring to releases of individuals with the DP of their eyes obscured differed statistically from that obtained in experiments conducted with intact animals (Fig. 5j vs k, U2(68, 34)  = 0.347, p < 0.005, Watson’s U2 test). The comparison among frequencies of radially—oriented sandhoppers under the blue filter and the vision of the sun prevented is statistically significant (G = 15.888, df = 2, p < 0.001, G test, Fig. 6d).

Discussion

Morphological investigations by microscopy

Our morphological investigations revealed that the rhabdom of T. saltator (Fig. 2c) is fused-type as found in other amphipods (Hallberg et al. 1980), and possesses cross sectional differences between DP and VP at 10 μm from the proximal end of the crystalline cone. The pioneering research carried out by Bertolazzi (1937) and Debaisieux (1944) on some species of gammarid amphipods (not T. saltator) show the existence of the fifth cell. This is a character shared by several species of amphipods living in different environments (freshwater, sea, deep sea, pelagic, benthic; Donner 1971; Ball 1977; Meyer-Rochow and Tiang 1979; Meyer-Rochow et al. 1991; Cohen and Frank 2006; Baldwin Fergus et al. 2015). However, Ercolini (1965) showed that the ommatidium of the compound eye of T. saltator possesses only four retinular cells. We found that there are only four rhabdomeres in the apical region of the rhabdom from the proximal end of the crystalline cone to about 5 μm depth of the rhabdom, but there are five retinular cells below 10 μm depth from the crystalline cone (Fig. 2 c–i). Ercolini probably observed at the apical region of the rhabdom of T. saltator.

The area ratio of the R5 rhabdomere to the whole rhabdomeres (R1–4 and R5) was larger in the DP and gradually smaller in the median part to VP (Fig. 2d–i). Our results indicate that the area occupied by R5 shows a regionalization in the compound eye of T. saltator.

We assumed that the structural specializations of the retina play a role of the functional differences between DP and VP, therefore, we estimated roughly the visual filed from the light microscopic data (Fig. 3). Since the number of the ommatidia of T. saltator increases with the growth of the body as reported for Parhyale hawaiensis (Ramos et al. 2019) and smaller ommatidial crystalline cones are frequently found at the peripheral region of the eye, we omitted these small ommatidia which possess small crystalline cones when we estimated the visual field. The visual field size in T. saltator was investigated using their pseudopupil (Beugnon et al. 1987); the individuals could potentially detect visual stimuli at 101° of vertical elevation. Our optical estimation of vertical elevation from the crystalline cone direction was wider, ca. 145°. Further experiments will be needed to confirm the origin of the visual field of T. saltator. The acceptance angle of each retinular cell measured by intracellular recordings in different parts of the eye might reveal the precise angle of each ommatidium.

Electrophysiological recordings

In electroretinogram experiments, both the waveforms of corneal negative- and positive responses were observed (Cohen et al. 2010), however, our data obtained from a proximal area were not stable showing both negative- and positive-responses. Therefore, we only used the corneal positive waveforms (Fig. 4b, insertion) which were obtained by the electrode placed on the surface layer of the crystalline cone (just below the cornea). In the V-log I curves from DP and VP which were obtained by the stimulation of white light, there were no apparent differences at the initial slope, but a slight difference at the saturation level (Fig. 4a). The spectral sensitivity curves obtained from DP and VP showed the apparent difference in shorter and longer regions (Fig. 4b). The spectral sensitivity curves obtained from DP showed two distinct maxima, one in the long-wave range peaking at 550 nm and one in the UV-blue range peaking at 430 nm, whereas the curves of VP possessed high sensitive region only around UV-blue and was decreased gradually in longer wavelength. Therefore, the spectral sensitivity curves indicate the existence of a regionalization. However, the relation between the electrophysiological data and morphological regionalization difference. We still don’t know the role of R5; however, in P. hawaiensis, the spectral absorption bands estimated from the genetic results differed between R1-4 and R5 (Ramos et al. 2019). A UV-blue sensitive pigment is thought to be present in distal R5 cells, while a green sensitive pigment is present in proximal R1–4 cells (Mezzetti and Scapini 1995; Cohen et al. 2010; Ramos et al. 2019). Behavioural studies suggest UV-blue sensitivity is essential for sun-compass behaviour, while longer-wavelength sensitivity is required for entrainment of endogenous activity rhythms (Ugolini et al. 1993; Forward et al. 2009a, b). Therefore, R5 cell of T. saltator also might have different colour absorption band. To confirm our speculation, we need further experiments such as the in situ experiment of T. saltator.

The insects which use skylight polarization for orientation or cruising-course control detect the polarized skylight pattern mediated by the ommatidia of the dorsal rim area (DRA), the average spectral sensitivity curve showed a prominent UV and/or blue response (Horváth and Varjù 2004; Dacke et al. 2011; Cronin et al. 2014). However, T. saltator does not use the skylight polarization pattern as a compass cue (Ugolini et al. 2013). It has been speculated to be responsible for facilitating the detection/use of other cues by enhancing the intensity/spectral contrast across the sky (Ugolini et al. 2013). The broad spectral sensitivity curve obtained from DP showed two distinct peaks at the blue and green regions indicate it possesses at least two different receptor types. The necessary basic equipment for colour vision is to have at least two types of photoreceptors that differ in spectral sensitivity locating in approximately the same area (Skorupski and Chittka. 2011), therefore, DP is thought to have colour vision. Comparing several absorbance spectra of visual pigments (Govardovskii et al. 2000), both the spectral sensitivity curves obtained from DP and VP were broader than the curves of absorbance spectra. The broad spectral sensitivity curves may include different photoreceptor cells corresponding to different wavelength bands, or maybe due to the effects of pigments such as ommochrome and eye structure. Moreover, the broader curve at the longer wavelength region obtained from VP also suggests the existence of retinular cells which possess λmax in a longer wavelength region. The future intracellular experiments and/or genetic sequencings of opsin genes and in situ experiments will clarify the broad spectral sensitivity curves obtained by ERG method.

Celestial compass experiments

In insects, dung beetles also can use spectral cues for orientation. Testing them with monochromatic (green and UV) light spots in an indoor arena, the results suggest it could potentially use the celestial chromatic gradient as a reference for orientation (el Jundi et al. 2015). Our data strongly suggest the high possibility that DP retina of T. saltator possesses colour discrimination ability for orientation behaviour (Ugolini et al. 2013).

It is known that the visual pigments of animals have a maximum peak of α-band and a smaller peak of β-band exists at the shorter wavelength (Stavenga et al. 1993). The spectral sensitivity curves of VP obtained by ERG method show a highly sensitive region around UV-blue and was decreased gradually in longer wavelengths. ERG recordings represent the overall spectral sensitivity of several ommatidia and, therefore, the summation of many retinular cells that were irradiated from a certain direction. It is difficult to find the peak in a longer wavelength region in the spectral sensitivity curve of VP, however, there must be the retinular cell peaking at longer wavelength region. Because of this reason, the VP region also might have colour discrimination ability to detect differences in the seaward and landward landscape.

Indeed, sandhoppers with the DP of their eyes painted met higher difficulties in orientating than the other groups of animals since in any experimental condition lower percentages of radially-orientated individuals were recorded and are more dispersed than intact individuals and animals with the VP of the eyes occluded (Fig. 5). Moreover, in experiments of spectral filtering with the vision of the sun prevented their mean orientation was also deflected by 55° with the respect that recorded for intact sandhoppers. In releases carried out with the vision of the sun precluded, the goodness of orientation of individuals with the upper area of their eyes painted reached fairly lower values than those recorded in releases of the other groups of individuals. Therefore, our investigations suggest that in T. saltator the dorsal 1/3 part is involved in the correct identification of celestial cues when the sun not visible. Instead, the rest of the eye seems not involved in the recognition of celestial orienting factors (but the sun) by this species. In fact, sandhoppers with the VP of their eyes obscured orientated in each experimental condition significantly toward their expected direction and frequencies of not radially-orientated lower than 40% were registered (Fig. 6).

Our paper suggests that both the DP and the VP of the eyes can identify the sun as a compass cue. Moreover, despite functionally and anatomically different dorsal and ventral regions, T. saltator does not use the celestial polarization pattern as a compass cue like some terrestrial and aquatic arthropods, reviewed by Horváth and Varjù (2004) and Cronin et al. (2014), respectively. It does instead rely on the intensity/spectral gradient across the sky (Ugolini et al. 2013) for which the two eye regions appear to be specialized. In the fiddler crabs, there are regularly stacking microvilli in alternating orthogonal bands within fused rhabdoms which fit to detect polarized light (Alkaladi et al. 2013). It is still possible to assume that T. saltator uses the perception of the polarized light in its orientation. In fact, both in DP and VP, the microvillar orientation of R1 and R3 are parallel, and that of R4 and R6 are also parallel, and this parallel microvillar orientation is orthogonal between the two pairs (Fig. 2). Therefore, we can speculate that the VP detects the polarization of the reflected light from the seawater surface and the pol-light perceived by the DP could increase the intensity and/or spectral gradients in the sky.

However, although a structural and functional eye regionalization has been pointed out in fiddler crabs species (Zeil and Al-Mutairi 1996; Smolka and Hemmi 2009; Alkaladi et al. 2013), no evidence has ever been provided on the involvement of specific eye regions in the celestial orientation of crustaceans. Our investigations on T. saltator indicate for the first time in Crustacea the existence of an area located in the DP of its eye specialized in detecting certain wavelength ranges involved in the perception of skylight cues.