Photoreceptors and diurnal variation in spectral sensitivity in the fiddler crab Gelasimus dampieri

To animals that can perceive colour, the natural world is bursting with colour signals. These signals are often vital to an animal's survival and fitness. We know that animals use colour signals to locate food sources (Chittka and Menzel, 1992), in mate choice (Hunt et al., 2001) and in visual communication (reviewed in Osorio and Vorobyev, 2008). How animals perceive these colours depends entirely on the colour vision system with which they are equipped. True colour vision, the ability to discriminate between different wavelengths of light, or different spectral radiance distributions, requires at least two distinct spectral channels. The range of colours that can be discriminated and the accuracy with which an animal can make colour discriminations generally increases with increasing spectral channels and the resulting dimensionality of colour space (Kelber et al., 2003).

Fiddler crabs are colourful animals that incorporate many colours in their exoskeletons, with patches on their carapace and appendages variously spanning the regions of human colour space that we categorise as blue, green, yellow, orange and red (Crane, 1975). These colours are linked to species recognition and intraspecific signalling (Detto, 2007; Detto et al., 2006; Detto and Backwell, 2009). For example, female Austruca (=Uca) mjoebergi (Shih et al., 2016) recognise conspecific males by the colours of their chelae, and prefer males with high UV reflections (Detto, 2007; Detto et al., 2006; Detto and Backwell, 2009). Nevertheless, no studies have provided evidence that fiddler crabs have the required number of spectral channels to discriminate between the various colours they exhibit or between different shades (whether hue, saturation or brightness) of similar colours, as would be required in mate choice (Horch et al., 2002; Hyatt, 1975; Scott and Mote, 1974). The importance of colour signals in fiddler crab behaviour has prompted investigations into the colour vision system of fiddler crabs (Alkaladi, 2008; Falkowski, 2017; Horch et al., 2002; Hyatt, 1975; Jordão et al., 2007; Rajkumar et al., 2010; Scott and Mote, 1974; Smolka, 2009). Yet, despite the considerable effort, some fundamental properties of how these animals see colour, including the number of spectral channels they possess, remain poorly understood. We know that fiddler crabs have a relatively broad spectral sensitivity, extending from around 400 to 600 nm, and are maximally sensitive to the region of the spectrum we describe as blue–green (Horch et al., 2002; Hyatt, 1975; Jordão et al., 2007; Scott and Mote, 1974). The width of their spectral sensitivity suggests that a few visual pigment types are contained within their eyes.

Like most crustaceans, ommatidia in the fiddler crab compound eye are composed of eight retinular cells. These are differentiated into proximal retinular cells (R1–R7) and a small distal retinular cell (R8) (Alkaladi and Zeil, 2014; Shaw and Stowe, 1982). Each rhabdom is surrounded by a dense layer of screening pigment which spectrally filters light passing down the rhabdom, hence altering the retinular cells’ effective spectral sensitivities (Alkaladi and Zeil, 2014; Jordão et al., 2007). Jordão et al. (2007) used microspectrophotometry (MSP) in an attempt to characterise the light absorption of the visual pigments located in the proximal retinular cells (R1–R7) and the screening pigments of four fiddler crab species: Leptuca pugilator, Minuca pugnax, Afruca tangeri and Gelasimus vomeris (Shih et al., 2016) (all of which were formerly assigned to the genus Uca). Their results suggest that a single visual pigment type, absorbing maximally between 525 and 600 nm, exists in R1–R7 in all four species. Owing to the small size of the distal R8, their study was not able to measure the visual pigment in these cells. Furthermore, the high density of screening pigments in fiddler crabs makes it difficult to measure the absorbance spectrum of their visual pigments without some spectral contamination from the surrounding screening pigments.

In contrast, molecular biological techniques have found evidence that at least two visual opsins are expressed in G. vomeris (Alkaladi, 2008) and L. pugilator (Rajkumar et al., 2010). In G. vomeris, two opsin genes (UcaOpsin1 and UcaOpsin2) are differentially expressed across the eye (Alkaladi, 2008). In L. pugilator the opsin genes UpRh1 and UpRh2 appear to be differentially expressed in R1–R7, with one of the retinular cells co-expressing the opsin genes (Rajkumar et al., 2010). In both species, these opsins are predicted to be medium-wavelength sensitive (MWS). In addition, Rajkumar et al. (2010) found in L. pugilator that R8 expresses the opsin gene UpRh3, which has a similar sequence to UV opsins in other arthropods. Both Rajkumar et al. (2010) and Alkaladi (2008), therefore imply that these species of fiddler crab have the molecular basis for at least dichromatic colour vision. However, these studies do not provide us with direct information about the classes of visual pigments nor the spectral sensitivity of the retinular cells, including the light filtering effects of the screening pigments.

Physiological measurements of fiddler crab spectral sensitivity are contradictory and a clear conclusion is still lacking. Three studies have measured the ERG of a number of species of fiddler crabs including M. pugnax and L. pugilator (Scott and Mote, 1974), and Leptuca (=Uca) thayeri, M. pugnax, L. pugilator and Minuca (=Uca) minax (Horch et al., 2002; Hyatt, 1975). Scott and Mote (1974) suggested that both M. pugnax and L. pugilator possess a single spectral type of retinular cell with a peak sensitivity ∼510 nm. However, their study suffered from several methodological limitations (reviewed in Horch et al., 2002). More recent ERG measurements (Horch et al., 2002; Hyatt, 1975) have implied that at least two distinct spectral types of retinular cells exist in fiddler crab eyes, one long wavelength sensitive (λ_max 500–540 nm) and one short wavelength sensitive (λ_max 430 nm) though neither study successfully provided conclusive evidence for this.

The discrepancies among all previous studies highlight the difficulty of conclusively defining the spectral types of retinular cells in the fiddler crab eye. Measuring spectral sensitivity intracellularly is difficult owing to the small size of the crab retinular cells, particularly in the case of R8 (Alkaladi and Zeil, 2014; Falkowski, 2017; Smolka, 2009). Additionally, there is some evidence that fiddler crabs possess a complex combination of visual pigments that are highly variable from cell to cell (Alkaladi, 2008; Falkowski, 2017; Rajkumar et al., 2010; Smolka, 2009), making it difficult to classify the spectral types of retinular cells unequivocally. If opsin co-expression does indeed exist within fiddler crab eyes, this is likely to lead to atypically broad spectral sensitivities and associated difficulties when estimating the contribution of different spectral classes of retinular cells to physiological data (Arikawa et al., 2003). Finally, physical and physiological changes within compound eyes can further complicate the matter. Physical changes observed in crustacean eyes include daily changes to opsin concentration within the rhabdom (Battelle, 2013; Katti et al., 2010) and daily movement of the light filtering screening pigments that surround the rhabdom (Olivo and Chrismer, 1980; Olivo and Larsen, 1978; Ribi, 1978; Satoh et al., 2017; Stowe, 1980). Both mechanisms can affect the spectral sensitivity of retinular cells. The movement of screening pigments leads to a shift in peak spectral sensitivity of up to 35 nm in the crayfish (Orconectes sp. and Procambarus sp.) (Goldsmith, 1978) and up to 40 nm in the moth Adoxophyes orana (Satoh et al., 2017). We also know that the size of the rhabdom can vary throughout the day in a number of related species of crabs (Nässel and Waterman, 1979; Rosenberg et al., 2000; Rosenberg and Langer, 2001; Stowe, 1982; Toh, 1987) and depending on opsin concentration, rhabdomeric size changes may result in variable spectral sensitivity over the day (Katti et al., 2010). Increases in rhabdom length would be predicted to broaden spectral sensitivities (Brindley, 1960), but it is possible for changes in rhabdom width to alter spectral sensitivity too. Diurnal changes in rhabdomeric width have recently been shown for the fiddler crab A. tangeri (Brodrick et al., 2020). If waveguide modes vary as a result, it is possible that the contribution of surrounding screening pigments to spectral rhabdomeric absorption could affect the cells’ overall spectral sensitivity, with smaller diameter cells predicted to be more affected than larger ones (Van Hateren, 1984). As a further complication, the location of screening pigments within the ommatidia also changes diurnally in the fiddler crab L. pugilator (Fielder et al., 1971; Fingerman, 1970; Reddy et al., 1997). Thus, diel shifts in fiddler crab spectral sensitivities may also be anticipated.

Here, we investigate the spectral sensitivity of the Australasian fiddler crab Gelasimus dampieri (Shih et al., 2016) (formerly Uca dampieri Crane 1975), using ERGs, single cell intracellular recordings made with sharp electrodes, and an experimental behavioural paradigm based on fiddler crab escape performance. The spectral sensitivity of several simultaneously stimulated ommatidia were measured with ERGs using a method of stimulation that provides an improved signal to noise ratio, with higher spectral resolution and range (down to 350 nm) (Ogawa et al., 2015). By making use of monochromatic adaptation lights during ERG recordings, we were able to suppress the response of specific spectral classes of retinular cells, revealing a distinct UV-sensitive retinular cell. Single cell recordings revealed broad blue–green sensitive retinular cells that appear to co-express at least two opsins. Both ERGs and single cell recordings showed shifts in spectral sensitivity, of ∼25 nm, towards longer wavelengths over the course of the day. Spectral sensitivity measurements based on the escape response of the crabs supports conclusions made from physiological spectral sensitivity data.

Fiddler crabs were collected in daylight at low tide from a mangrove habitat near Learmonth (22°18S, 114°9E), south of Exmouth, Western Australia, in May 2014, April 2015 and July 2017, and from a mangrove habitat near Broome (17°57S, 122°14E) in July 2016. After field collection, crabs were housed in plastic containers with sea water and were transported by car to the University of Western Australia, Perth, WA. There they were housed in an artificial mudflat enclosure with an immersion/emersion tidal cycle in which crabs were immersed in sea water twice daily, under a 12 h light:12 h dark cycle (light phase from 06:00 h to 18:00 h) and were fed dried flake food (Aqua One Tropical Flakes) every 2 days. Selective adaptation experiments were conducted on four male and eight female crabs from the Exmouth collection site. ERGs measuring diurnal changes were conducted on four females and four males from the collection site in Broome. Single cell recordings were made from two male and four female crabs from the collection site in Exmouth. Behavioural experiments were conducted on two males and two females from the Exmouth collection site. Different groups of animals were used for the four different experiments (two ERG experiments, one intracellular, and one behavioural experiment). No single animal was used across two different experiments.

Fiddler crabs were treated according to UWA Animal Ethics Committee (AEC) approved methods (UWA AEC project numbers RA/3/100/1515 and RA/3/400/1020). A fiddler crab was removed from the holding facility and its carapace was glued with Loctite Superglue (ethyl cyanoacrylate) to a small insulated plastic disc that was then glued to a vertical metal rod so that the crab was in its natural posture. Both eyes were stabilised in a natural upright position by gluing the eye stalks to a wooden post as depicted in Fig. 1. The crab was then partially submerged in a plastic container filled with sea water and placed in a grounded Faraday cage.

To record the electrophysiological response of the eye to the light stimuli, the ERG was recorded and used to determine the spectral sensitivity of the illuminated equatorial and frontal part of the eye. A 0.254 mm diameter platinum electrode was placed on the cornea of the stimulated compound eye with conductive gel (Livingstone International, New South Wales, Australia). A shielded silver/silver chloride pellet was placed inside the sea water bath containing the crab and served as the indifferent electrode. The wire placed on the stimulated eye acted as the recording electrode. ERGs were recorded through a differential amplifier (DAM50, World Precision Instruments, FL, USA) connected to a computer via a 16-bit data acquisition card (USB-6353, National Instruments, Austin, TX, USA), with the ground connected to the recording table. Data were acquired at 5 kHz using custom made software in MATLAB 2015B (MathWorks) and the signal was monitored using an oscilloscope (Tektronix 5223, Tektronix, Beaverton, OR, USA). Measurements were taken across the ultraviolet to visible (UV–VIS) spectrum, taken in sequence from 610 to 350 nm in 20 nm steps and then from 360 to 620 nm in 20 nm steps.

The spectral sensitivities of single retinular cells were determined by intracellular measurements with sharp electrodes. A small incision was made in the cornea at the top of the eye using a razor blade, to provide access to the retinular cells. Microelectrodes (borosilicate glass World Precision Instruments, Inc.; 1.2 mm outer diameter, 0.68 mm inner diameter) were pulled on a laser-based micropipette puller (P-2000, Sutter Instrument, Novato, CA, USA) yielding tip resistances of 40–80 MΩ when filled with 1 mol l⁻¹ potassium chloride. The electrode was then inserted into the retina through the small hole and served as the recording electrode. The shielded silver/silver chloride pellet in the sea water bath served as the indifferent electrode. Membrane potentials were recorded via an amplifier (Getting Model 5A, Getting Instruments, San Diego, CA, USA), connected to a computer via the 16-bit data acquisition card (see above). The successful penetration of a retinular cell was indicated by the amplitude of the response to a flash of light. As a criterion for a healthy recording, the response amplitude to a flash of light needed to exceed 15 mV. After penetrating a retinular cell with the electrode, the position of the optical stimulation fibre was adjusted to produce the maximum retinular cell response. Measurements were taken in sequence across the UV–VIS spectrum from 610 to 350 nm in 20 nm steps and then from 360 to 620 nm in 20 nm steps. Crabs were semi-immersed in sea water for the duration of the experiment. Crabs were euthanized at the conclusion of each intracellular recording.

Monochromatic stimulation and adaptation lights were produced by computer-controlled monochromators (TILL Polychrome V, Till Photonics, Gräfelfing, Germany) illuminated with 150 W xenon lamps (USHIO Xenon Short Arc Lamp UXL-150S, Ushio America, CA, USA). White light, with a range from 185 to 2000 nm (Fig. S1), was provided by a 35 W xenon light source (HPX-2000, Ocean Optics, FL, USA). Our light stimulation protocol is similar to that described by Ogawa et al. (2015). To investigate visual spectral sensitivities a flickering light stimulus was used, alternating with a 50% duty cycle between light and dark at a frequency of 10 Hz. The light phase alternated between monochromatic and white light illumination. The angular size of the circular stimulus (aperture of the lens) was approximately 1.5 deg in diameter when viewed by a crab positioned at the centre of the recording setup (15 cm away from the light). Therefore, based on estimated interommatidial angles (1.25 deg and 0.26 deg, horizontal and vertical, respectively) in G. dampieri (Bagheri et al., 2020), approximately 8–12 ommatidia or 64–72 retinular cells viewed at least part of the aperture when the centre of the eye was stimulated. Examples of a raw ERG recording and a filtered intracellular recording are shown in Fig. S2. The white light intensity was invariant but, at each wavelength tested, the intensity of the monochromatic light was adjusted to produce the same electrophysiologically measured response amplitude from the monochromatic light to dark transition as that produced by the white light to dark transition. By comparing the photon irradiance of monochromatic light, of different wavelengths, required to match the photon irradiance of white light allowed us to calculate the crab's relative spectral sensitivity for all measured wavelengths. Additionally, we calculated the mean amplitude of the response to the standard white light to dark transition throughout each recording to provide a way of comparing response amplitudes across crabs.

Two different ERG experiments were conducted to investigate the spectral types of retinular cells contributing to the spectral sensitivity of G. dampieri. In the ‘adaptation’ experiments, the eye was illuminated with an additional monochromatic adaptation light to reduce the contribution to the measured signal from some retinular cell classes by reducing their effective stimulus contrast. Adaptation lights were delivered through a beam splitter to achieve exact spatial alignment with the stimulus light. Adaptation lights had a bandwidth of 15 nm (full width at half maximum transmission) and peaked at 420 nm, 430 nm, or 530 nm. Irradiance measurements of the adaptation lights and the invariant white light were made using an ILT1700 radiometer and SED033 detector (International Light Technologies Inc., MA, USA). The brightest possible irradiance of each adaptation light was used in all adaptation experiments. Eight female and four male crabs were used; however, not all crabs were measured in all conditions. In the ‘time of day’ experiments, we examined whether the spectral sensitivity of the crabs varied predictably across the day. Spectral sensitivities were measured by recording ERGs at four different times during the day: 09:00 h, 12:00 h, 15:00 h and 18:00 h. Using a Latin square design, the spectral sensitivity of each fiddler crab was measured at all four times in a pseudo-random order. We measured four females and four males, and all crabs were measured at each time point. Measurement duration was approximately 20 min and the average stimulus intensity remained constant at each time point.

The absorptance spectrum was fitted to the scaled spectral sensitivity data from physiological recordings over the wavelength range 400 nm to 600 nm owing to the lack of data below 400 nm in the screening pigment template. By minimizing the root mean square difference between the modelled data and the ERG recordings we were able to determine the best-fitting visual pigment λ_max values, their relative occupancies (relative contribution) (p_i) and the peak absorbance coefficient of the screening pigment (k_s,max). To explore certain hypotheses these parameters could be constrained or unconstrained. As previously mentioned, the rhabdom of fiddler crabs is surrounded by a dense layer of screening pigment (Alkaladi and Zeil, 2014; Jordão et al., 2007), therefore, when fitting the optical model to our physiological data, a screening pigment template was always included in the model. The exclusion or inclusion of the screening pigment template did not greatly affect the model result (Supplementary Material, Fig. S3). When comparing models with the same number of unconstrained parameters we compared the sum of squares between the models to identify the best-fitting model. When comparing models with different numbers of unconstrained parameters an F-test was performed to test the level of improvement in the fit between models (Anderson and Conder, 2011). Models were fitted to the data of ERG experiments independently as the crabs used in each experiment came from geographically distinct populations separated by ∼1200 km. The intracellular data came from crabs from the same population as the ‘selective adaptation’ experiment, models were again fitted to these data sets independently because of the difference in methodologies.

The behavioural apparatus consisted of a treadmill setup similar to that described by How et al. (2012). A 100 mm diameter polystyrene ball balanced on a cushion of air flowing from a reservoir was used as the treadmill. The crab was fixed in position on top of the treadmill using a small metal rod fixed to the carapace of the crab. The rod was attached to a pivot system allowing the crab to rotate about the vertical axis and to walk freely in any direction along the surface of the treadmill. The treadmill was placed in the middle of a testing arena (53.5×53.5 cm) where the floor was level with the top of the treadmill, which was surrounded by three liquid crystal display (LCD) monitors (33×51 cm; DELL 90-130, Ireland) and one white diffusing screen. LCD monitors showed a stationary image, simulating a mudflat with a dark ground and a bright sky, providing a steady but dim background light. The stimulus light was delivered from a computer controlled monochromator (as above) via an optical fibre and imaged through a lens before passing through a neutral density (ND) filter, a motorised aperture, and finally projected onto the white diffusing screen from a distance of ∼920 mm. The background of the white diffusing screen was not illuminated. The resulting stimulus was an expanding, circular monochromatic light, which mimicked a sphere approaching the crab along a straight line on a direct collision course. The motorised aperture, was controlled by a stepper motor. The angular size of the stimulus exponentially increased from 2.1 deg to 23.9 deg over 3 s.

Each animal was mounted on the treadmill and allowed 3–5 min to acclimate. A stimulus of a given wavelength and intensity was then presented and the crab's response recorded. Crab behaviour and stimulus events were recorded using a high definition (HD) digital (1920×1080 pixels) video camera (Handycam HDR-CX550, SONY, Japan) recording at 25 fps located directly above the testing arena. To prevent habituation, each crab was presented with only three trials every 2 days (Tomsic et al., 1998). Video recordings were analysed after each trial to determine if stimuli resulted in an escape response. Test wavelengths included: 360 nm, 380 nm, 400 nm, 420 nm, 440 nm, 460 nm, 480 nm, 500 nm, 540 nm and 600 nm. Each crab was randomly assigned an order of test wavelengths.

The ND filters inserted into the light path and the exit slit of the monochromator were used to change the intensity of the stimulation. The stimulus radiance (μmol m⁻² sr⁻¹) was measured using an ILT1700 light spectrometer (International Light Technologies Inc., MA, USA) that was positioned at the crabs’ location in front of the diffusing screen. At each wavelength the maximum intensity was recorded. Monochromator exit slits and ND filter combinations were found to produce decreasing intensities in a logarithmic fashion with a step size of 25% and 19 possible steps beginning from step one (100% intensity) and decreasing to step 19 (0.56% intensity). A modified staircase procedure was used to determine the minimum stimulus intensity needed to elicit an escape response. For each test wavelength the staircase procedure began with step one and if the crab responded within four presentations the intensity was decreased by four steps (31.64% intensity) and the stimulus was re-presented. Intensities were decreased in steps of four until the crab failed to respond to four presentations of the same intensity. Once this occurred, the intensity was then increased by two steps and depending on the response outcome the intensity was either increased or decreased by one step. This procedure would continue until a one-step intensity difference resulted in a different response outcome. The higher intensity of the one-step intensity difference was reported as the threshold intensity for that wavelength. Using the inverse of the threshold values we then produced a relative spectral sensitivity curve. If the full intensity was presented four times with no response, we concluded the animal was not sensitive to that wavelength and the relative spectral sensitivity for that wavelength was zero. Control stimulations, where the light was blocked by a black plastic shield, were presented at regular intervals (every fifth presentation) to ensure potential unintended stimuli, such as any vibrations from the motorised aperture, were not affecting the response.

Electroretinogram measurements in G. dampieri showed relatively broad spectral sensitivities peaking between 420 and 460 nm (Fig. 2). There was no difference between the sensitivities of male and female crabs, therefore both sexes were combined in the mean data. Optical models including two visual pigments (Fig. 2B) fit the ERG recordings significantly better than models including a single visual pigment (Fig. 2A) (F=54.36, P<0.05). The best-fitting model to the mean ERG data (Fig. 2B) includes visual pigments with λ_max values of 431 nm (VP1) and 482 nm (VP2), with relative occupancies (or contributions) (p_i) of 0.53 and 0.47, respectively. The screening pigment absorbance coefficient in this model was 0.07 (Fig. 2B).

To separate the responses of distinct spectral classes of retinular cells, adaptation lights were used concurrently with the stimulation lights. When adapted with 420 nm or 430 nm light, the peak spectral response of the fiddler crab eye shifted from ∼440 nm to between 350 nm and 360 nm (Fig. 2C). An increase in relative sensitivity was seen from 350 nm to 390 nm under both 420 nm and 430 nm adaptation conditions; however, the relative sensitivity from 400 nm to 600 nm was largely unchanged between the 420 nm, 430 nm and no adaptation conditions. Adding a 530 nm adaptation light did not produce a shift in peak sensitivity and only slightly decreased the sensitivity around 500 nm of the stimulated ommatidia (Fig. 2D).

To further investigate the effect of each adaptation light, we fitted the model from Fig. 2B to the spectral responses of each adaptation condition, leaving the λ_max of each visual pigment and the screening pigment absorbance coefficient constrained, but allowing the relative contribution of the visual pigments to vary. The λ_max of each visual pigment was constrained to 431 nm for VP1 and 482 nm for VP2. The screening pigment absorbance coefficient was constrained to 0.07. Despite the strong adaptation lights, the relative contribution of the two visual pigments varied only slightly (between 5 and 15%) compared to the no adaptation light condition (Table 1).

Calculated contrast values show that the contrast ‘seen’ by putative visual pigments would have varied very strongly under the different adaptation conditions if they were functionally independent (Table 2), but the model-estimated visual pigment contributions remained almost completely stable as seen in Table 1. In other arthropods minimum contrast thresholds to elicit a response from a retinular cell are between 2.5 and 6% (O'Carroll and Wiederman, 2014; Ogawa et al., 2019; Ryan et al., 2020) and in G. dampieri it is predicted to be ∼3% (Z.M.B., unpublished data). If we assume each visual pigment is contained within a discrete retinular cell, then we would expect to see no response to the 10 Hz stimulation from the retinular cells under adaptation conditions where the contrast ‘seen’ was below 2.5%, leading to a significant change in the relative contributions of modelled visual pigments. In all adaptation conditions, however, this was not the case, and modelled relative contributions remained similar to the no adaptation model. Even though the recorded spectral sensitivity remained mostly unchanged, we have clear evidence that the adaptation lights were visible to the crabs. The 420 nm, 430 nm and 530 nm adaptation lights reduced the amplitudes of responses to the standard white light to dark transition to (respectively) 48.9%, 52.5% and 74.6% of the amplitude recorded under conditions of no adaptation. In addition, the strong UV component of the spectral response unveiled under 420 nm and 430 nm adaptation lights (Fig. 2C) clearly highlights the effectiveness of these shortwave adaptation lights. Under 530 nm adaptation the strong UV response component was not seen. If we assume the 431 nm and 482 nm visual pigments are contained within the same retinular cell, then we may expect this to be the case. For this putative retinular cell, the calculated perceived contrast was twice as high under the 530 nm adaptation condition compared to the other adaptation conditions, indicating that the adaptation effect of the 530 nm light was weaker compared to the other adaptation conditions.

Electroretinogram recordings were taken from eight G. dampieri, four females and four males, at four different times of day: 09:00 h, 12:00 h, 15:00 h and 18:00 h. Fig. 3A displays the mean ERG data recorded at each time, fitted with optical models. A clear shift in spectral sensitivities towards longer wavelengths later in the day can be seen and this is quantified by the optical model results in Table 3. The peak sensitivity at 09:00 h shifted by ∼25 nm towards longer wavelengths by 18:00 h.

In order to understand the possible causes of this shift, three different hypotheses were compared: (1) the migration of the screening pigments that surround the rhabdom is causing the shift; (2) differential regulation of opsin co-expression is causing the shift; and (3) the combined effect of the migration of screening pigments and the differential regulation of opsin co-expression are causing the shift. To test these three hypotheses, the ERG data recorded at different times of day were combined and fitted with a model that best fitted all data, as shown in Fig. 3B. This model included two visual pigments with λ_max values of 454 nm (VP1) and 496 nm (VP2) and relative contributions of 0.73 and 0.27, respectively, with a screening pigment having an absorbance coefficient of 0.005. This model was then fitted to the ERG data from each time of day separately with the following constraints: (1) (screening pigment migration) the screening pigment absorbance coefficient (k_s,max) was unconstrained but all other parameters were constrained to the results from the best-fitting model; (2) (opsin regulation) the relative contributions of visual pigments (p_i) were unconstrained with all other parameters constrained; and (3) (combination) all parameters were unconstrained except the λ_max for each visual pigment. The sum of squares between the data and the models were compared between each hypothesis (Fig. 3C); however, no clear difference was found. The ERG data fit with each model and the best-fitting parameters for each model are shown in Fig. S4 and Table S1.

The spectral sensitivities of 19 retinular cells were measured intracellularly using sharp electrodes between 12:00 h and 19:00 h (Fig. 4A). During one single cell recording (cell 19), we were able to hold the cell over ∼3 h, which permitted measurement of spectral sensitivity from the same cell at two different time points: 13:00 h and 15:00 h (Fig. 4B). A clear shift in spectral sensitivity towards longer wavelengths can be seen in this cell, with the peak shifting ∼20 nm and the width of curve increasing during the ∼3 h time period. Of the remaining 18 recordings, there were two distinct temporal groups, with one group of cells recorded between 12:00 h and 14:00 h (early cells, N=5) (Fig. 4C) and a second group between 15:00 h and 19:00 h (late cells, N=13) (Fig. 4E). The peaks of the cells recorded from later in the day were shifted towards longer wavelengths by ∼20 nm compared to the peaks of cells recorded from earlier in the day. All response curves were broad and had similar widths and peaks, depending on time, to the ERG-measured response curves.

A model including two visual pigments and a screening pigment was fitted to the spectral responses of retinular cells measured early (Fig. 4C), late (Fig. 4E) and to the responses from Cell 19 at each time point (Fig. 4D,F). Visual pigment contributions and the screening pigment absorbance coefficients were left unconstrained. The best-fitting model parameters show that the relative contributions (p_i) of the 486 nm visual pigment increased with time in both data sets (Table 4).

During the behavioural spectral sensitivity experiment, an escape response was recorded when a crab ran away from the looming stimulus. Of the trials where an escape run was elicited, crabs consistently sprinted directly away from the stimulus (Fig. 5A, Rayleigh test, N=24, P<0.01). The spectral sensitivity estimate from this experiment has a peak sensitivity of ∼460 nm and decreases both towards shorter and longer wavelengths (Fig. 5B). Crabs were tested between 09:00 h and 17:00 h; however, as a result of the experimental design and issues relating to habituation, it was not possible to test the effect of time of day on spectral sensitivity.

Our results show that the fiddler crab G. dampieri possesses at least two spectrally distinct types of retinular cells, with peak sensitivities in the UV and blue–green region of the spectrum. The blue–green sensitive retinular cells appear to contain at least two visual pigments and the measured spectral sensitivity shifts to longer wavelengths over the course of the day. Preliminary spectral sensitivity estimates based on the fiddler crabs' escape behaviour support the physiological spectral sensitivity measurements.

The finding of a UV-sensitive retinular cell in our selective adaptation experiments (Fig. 2C) is consistent with two previous studies. Rajkumar et al. (2010) provided evidence that L. pugilator expresses an apparent UV-sensitive (UVS) opsin in R8, and Detto and Backwell (2009) showed that the fiddler crab, A. mjoebergi can use UV cues in mate choice. The preliminary results from our behavioural experiment also support this finding, with two crabs responding to test wavelengths of 360 nm and 380 nm. We were unable to record intracellularly from the UV-sensitive retinular cell and therefore cannot be sure about the exact wavelength of its peak sensitivity, but we estimate it is shorter than 360 nm. The results provided in our selective adaption experiments represent the first physiological evidence for a UV-sensitive retinular cell in fiddler crabs.

The widths of the spectral sensitivity curves from both our ERG and single cell recordings suggest the blue–green wavelength sensitive retinular cells contain at least two visual pigment types and this is supported by the optical modelling (Figs 2 and 4). An important consideration is that the screening pigment absorbance template included in our modelling was measured from G. vomeris and not G. dampieri (Jordão et al., 2007). However, the two species are very closely related (Shih et al., 2016) and the presence or absence of the screening pigment did not change the conclusion that at least two visual pigment types are contained within the blue–green sensitive retinular cells (e.g. Fig. S3). It is possible, though, that the template used in our modelling is not representative of the spectral absorption of G. dampieri screening pigment and such measurements should be considered in future research. A second piece of evidence for a two visual pigment retinular cell comes from our ERG recordings, which show that the spectral responses to wavelengths between 400 and 600 nm remained mostly unchanged under all adaptation conditions. Given that the effective contrast calculated for the putative VP2 (482 nm) under all adaptation conditions was below the threshold contrast for most arthropods (Table 2), we would have expected a significant reduction of responses to wavelengths between 500 and 600 nm if VP2 and VP1 were contained within distinct retinular cells. This was not the case, suggesting the effects of the adapting lights were either overestimated or that two visual pigments are contained within a single retinular cell. The adapting lights did not completely suppress the response of the main population of retinular cells, but we saw a clear effect of the adapting lights on the amplitude of the ERGs and the 420–430 nm adapting lights were strong enough to unveil the UV-sensitive retinular cell (Fig. 2C). Why we did not see a strong increase in the UV response under 530 nm adaptation may be explained by the effective contrast ‘seen’ by a putative retinular cell containing two visual pigments under 530 nm adaptation. The contrast seen under 530 nm adaptation was double that under 420 nm and 430 nm adaptation. The results from Ogawa et al. (2015), which used the same equipment, show clear effects of different adaptation lights on the spectral responses from Australian bull ants Myrmecia vindex and Myrmecia croslandi. Specifically, when using a 560 nm adaptation light, the spectral responses between 450 and 600 nm of these ants were drastically reduced (Ogawa et al., 2015). In contrast, in our 530 nm adaptation recordings, the reduction in spectral responses was minimal. Furthermore, previous ERG studies also reported no significant difference in spectral response curves when adapting lights between 500 nm and 600 nm were applied (Horch et al., 2002; Hyatt, 1975; Scott and Mote, 1974). Therefore, it seems likely that co-expression of at least two opsins within the blue–green retinular cells explains these results. These findings corroborate some of the results from the molecular biological studies of Alkaladi (2008) and Rajkumar et al. (2010), who found co-expression of MWS opsins in some, but not all, retinular cells. If there were a small number of pure retinular cells, such an arrangement could possibly explain the subtle shifts in our data. However, a simpler explanation of the subtle effects of the adaptation lights would be small differences in the ratio of the two opsins in different cells. In this situation, adaptation lights would change the contribution of those cells to the final recording, thereby leading to small shifts in the sensitivity (Fig. 2).

Our modelling suggests there is a combination of at least two different MWS visual pigments that underlie sensitivity between 400 and 600 nm and that it may depend on the population of crabs that was measured. For crabs collected from the Exmouth field site, one MWS visual pigment has a λ_max between 431 nm and 438 nm, and a second MWS visual pigment has a λ_max between 482 nm and 486 nm. For crabs collected from the Broome field site, the λ_max values were longer with one MWS visual pigment λ_max of 454 nm and the other 498 nm. Broome and Exmouth are separated by a considerable distance of ∼1200 km. The differences between populations were an interesting result and need to be further investigated. On average, across all experiments where spectral sensitivity was measured at ∼12:00 h the λ_max values of the modelled MWS visual pigments were 438 nm and 484 nm. Our visual pigment λ_max estimates are roughly congruent with the predictions from Horch et al. (2002) who suggested the presence of two visual pigments: one with a λ_max near 430 nm and the other with a λ_max between 500 nm and 540 nm in L. thayeri. Furthermore, the peak sensitivity and widths of the spectral response curves from Horch et al. (2002) are very similar to our spectral response curves from later in the day. The MWS visual pigment from Horch et al. (2002) has a slightly longer peak wavelength than our predictions, but their study used a different species of fiddler crab and did not account for lateral screening pigment effects.

In the present study, we found diurnal shifts in the spectral sensitivity of the fiddler crab, G. dampieri, in both ERG (Fig. 3) and intracellular recordings (Fig. 4). These shifts appear to be endogenous as the adaptation state of the crab remained the same throughout the measurement period. Using modelling, we have attempted to explain the cause of this shift (Fig. S4 and Table S1); however, a conclusive explanation remains elusive (Fig. 3C). There are a number of possible explanations for this shift. In this study we focused on the migration of screening pigments that has been shown to occur in L. pugilator (Fielder et al., 1971; Fingerman, 1970; Reddy et al., 1997) and differential regulation of opsin gene expression (DeLeo and Bracken–Grissom, 2020; Katti et al., 2010). Although there is no evidence that the latter occurs in fiddler crabs, there is evidence to suggest that the rhabdom width of the fiddler crab A. tangeri does fluctuate diurnally (Brodrick et al., 2020). The possibility of a shift in spectral sensitivity due to A1 to A2 chromophore lability was not modelled in our study as no brachyuran crab has been shown to utilise A2-based porphyropsin visual pigments (Cronin and Forward, 1988). To address the question of what is causing the shift in spectral sensitivity observed in G. dampieri, further research is needed that is beyond the scope of this study.

Our findings support the idea that fiddler crabs possess the minimum requirement for colour vision: two spectrally distinct retinular cells, containing three different visual pigments, one UVS (≤360 nm) and two MWS (∼438 nm and ∼484 nm). Both Detto (2007) and Hyatt (1975) demonstrated colour vision in fiddler crabs and showed that they are able to discriminate between two objects based only upon differences in the spectral distribution of their reflections and we have now provided the physiological basis for their findings. With this colour vision system, G. dampieri would be able to distinguish UV reflections from almost any other background reflection. Specular reflectance measurements have revealed that the white and blue parts of the G. vomeris cuticle are highly reflective in the UV region of the spectrum (Hemmi et al., 2006; Zeil and Hofmann, 2001). Very similar patterns of colours are seen in G. dampieri (Crane, 1975; Shih et al., 2016) and to a UV/blue–green colour vision system the blue and white patches provide strong contrast signals against the dark mudflat background (Zeil and Hofmann, 2001). These signals would increase the conspicuousness of conspecifics, allowing G. dampieri to more easily detect conspecifics for a range of different functions (Zeil and Hofmann, 2001).

Surprisingly, our results, like the results of all previous studies (e.g. Horch et al., 2002; Hyatt, 1975; Jordão et al., 2007 and Rajkumar et al., 2010) suggest that, despite the frequent display of yellow, orange and red colours (Crane, 1975), fiddler crabs do not possess two distinct spectral channels that would easily discriminate between these colours, which differ most in the longwave components of their spectral reflections. Notably, such patterns of colours on the G. dampieri carapace and claws reflect in the orange/red region of the spectrum. Furthermore, the carapace colours of other species such as Tubuca (=Uca) flammula are sometimes exclusively red and black. It is possible that small variations in opsin ratios between cells, or a small subset of retinular cells with narrow band spectral sensitivity, could provide these animals with a limited ability to discriminate between some long-wavelength colours if the signals from these retinular cells were kept separated for higher-level neural processing. Such small variations would also explain the very small shifts we have observed with our adaptation lights (Fig. 2). The apparent discrepancy between the colour signals displayed by fiddler crabs and their colour vision system offers an interesting challenge to better understand the evolution of colour signals and colour vision systems.

We would like to thank Nicolas Nagloo and Marcin Falkowski for their contributions in designing and constructing the equipment used to produce the visual stimuli. We would also like to thank the two anonymous reviewers for their thoughtful and insightful comments on the manuscript.