A combination of behavioural and electrophysiological experiments have previously shown that two species of stomatopod, Odontodactylus scyllarus and Gonodactylaceus falcatus, can differentiate between left- and right-handed circularly polarized light (CPL), and between CPL and linearly polarized light (LPL). It remains unknown if these visual abilities are common across all stomatopod species, and if so, how circular polarization sensitivity may vary between and within species. A subsection of the midband, a specialized region of stomatopod eyes, contains distally placed photoreceptor cells, termed R8 (retinular cell number 8). These cells are specifically built with unidirectional microvilli and appear to be angled precisely to convert CPL into LPL. They are mostly quarter-wave retarders for human visible light (400–700 nm), as well as being ultraviolet-sensitive linear polarization detectors. The effectiveness of the R8 cells in this role is determined by their geometric and optical properties. In particular, the length and birefringence of the R8 cells are crucial for retardation efficiency. Here, our comparative studies show that most species investigated have the theoretical ability to convert CPL into LPL, such that the handedness of an incoming circular reflection or signal could be discriminated. One species, Haptosquilla trispinosa, shows less than quarter-wave retardance. Whilst some species are known to produce circularly polarized reflections (some Odontodactylus species and G. falcatus, for example), others do not, so a variety of functions for this ability are worth considering.
Various aspects of vision are determined by environmental and behavioural need, including overall sensitivity, spatial resolution, colour sensitivity and in some species, polarization sensitivity (PS). PS is common among invertebrates (Waterman and Horch, 1966; Schwind, 1991; Labhart and Meyer, 1999), but also found in vertebrates such as fish (Hawryshyn, 1992, 2003; Coughlin and Hawryshyn, 1995; Roberts et al., 2004; Roberts and Needham, 2007). As a source of visual information and as a visual cue, the polarization of light is used by many insects for navigation (Rossel, 1993; Labhart and Meyer, 2002; Dacke, 2003; Wehner and Muller, 2006) and water body location (Schwind, 1984; Horváth and Varju, 1997). Other invertebrates such as cephalopods and crustaceans also use polarization to increase visual contrast (Temple et al., 2012; How and Marshall, 2014; How et al., 2015) and for visual signalling (Cronin et al., 2009; Marshall et al., 2014).
Polarization is a fundamental property of light and, in the context of animal vision, polarization refers to three measureable quantities for light that is made up of multiple waves (Goldstein, 2010): (1) the angle of polarization, (2) ellipticity and (3) the degree of polarization.
(1) The angle of polarization (AoP) defines the average angle at which the individual waves of light oscillate. If a linear polarizer is placed in the path of the light, then this is the angle at which the maximum intensity is transmitted.
(2) The electric fields of individual waves may be pictured as oscillating in a single plane or rotate as a circle or an ellipse around their direction of travel. Linearly polarized light (LPL) occurs when the multiple waves within a beam of light waves oscillate, on average, in a single plane. This contrasts with circularly polarized light (CPL), where the rotating individual waves result, at any one point in time, in an equal measureable intensity in all directions. Elliptically polarized light (EPL) is the most general form, in between LPL and CPL. As a numerical value, ellipticity ranges from −1 to +1 for left-handed CPL to right-handed CPL, respectively. A value in between represents EPL, and 0 is the special case for LPL.
An optical structure or device that creates a phase retardation of a quarter of a wavelength is called a quarter-wave retardation plate, or quarter-wave retarder. Materials that have the correct thickness and birefringence to convert LPL to CPL also work the opposite way, converting CPL to LPL.
(3) The degree of, or percentage, polarization is the ratio of the (averaged) intensity of the polarized portion of the light to its total (averaged) intensity. For LPL, this is the extent to which multiple waves oscillate at the same angle of polarization. Man-made polarizing filters generally achieve a percentage polarization close to 100%. However, in nature, the percentage polarization of light is generally much lower. Light in the sky that scatters in the atmosphere has a maximum of around 60% (Wang et al., 2016) and can be around 40% in the ocean (Cronin et al., 2003a). Polarized animal signals can have a percentage polarization of up to 80% (Marshall et al., 2014; Jordan et al., 2016).
Sensitivity to the polarization of light relies on the intrinsic PS of the retinal photoreceptors (Roberts et al., 2011). The rhabdomeric photoreceptors of invertebrates, particularly insects and crustaceans, are polarization sensitive due to their orientational order and unidirectional microvilli (Snyder, 1973; Snyder et al., 1973; Roberts et al., 2011). Many crustaceans arrange their microvilli in two interdigitating perpendicular directions and position their photoreceptors relative to the outside world, allowing for maximal sensitivity to horizontal (H) and vertical (V) polarized light (Waterman and Horch, 1966; Alkaladi et al., 2013; Marshall and Cronin, 2014). Stomatopods, benthic marine crustaceans, have a more complex retinal structure and organization of photoreceptors (Fig. 1A,B), which enables different information channels to be utilized in parallel (Marshall et al., 1991; Marshall and Cronin, 2014). The ommatidia in the dorsal and ventral hemispheres of stomatopod eyes follow similar organization seen in other crustaceans for sensitivity to linear polarization, except that the groups of microvilli in each hemisphere are oriented at ±45 deg with respect to each other (Marshall et al., 1991) (Fig. 1C). Furthermore, stomatopods exhibit a range of eye movements including rotation and scanning, and move their eyes both independently and asymmetrically. The ability to rotate their eyes around the eye-stalk axis makes the actual angle of PS relative to the world arbitrary, potentially allowing serial analysis of polarized light (Marshall et al., 1991) or specific optimization of PS for improving contrast of an object against a background (Daly et al., 2016).
Stomatopods use both LPL and CPL visual signals for inter- and intraspecific communication. Many species broadcast a variety of different linearly polarized visual signals (Cronin et al., 2003b,c, 2009; Chiou et al., 2005). For example, the first maxillipeds of many haptosquillids are involved in sexual signals (Chiou et al., 2011) and recently, How et al. (2014) showed that the linearly polarized dimension of this signal had evolved through a mechanism of sensory bias. Circular polarization signalling may also facilitate communication (Chiou et al., 2008) and constitute a private communication channel for stomatopods, one that remains invisible to other animals, but highly salient to conspecifics (Gagnon et al., 2015). So far, however, only a few species have been found with CPL reflections and there are some species that appear to lack any polarization signal at all, potentially relying only on colour. It is worth noting there are also several potential functions other than signalling such as contrast enhancement (Daly et al., 2016) or haze reduction (Schechner et al., 2003) that might make it worthwhile evolving sensitivity to one or both types of polarized light. One of the purposes of the study described here was to begin to clarify which species have the potential sensory mechanism to detect CPL.
Stomatopod sensitivity to CPL relies on what are known as the R8 cells (retinular cell number 8) in the row 5 and 6 receptors of the midband photoreceptors (Fig. 1) (Chiou et al., 2008). The R8 cell occupies a position above the main R1–7 rhabdom and acts as a quarter-wave retarder, converting any incoming CPL to LPL. The result is that the underlying R1–7 rhabdom becomes effectively sensitive to CPL by the fact that they detect the converted LPL. What is particularly unusual about the R8 cell is that the retardation is wavelength insensitive (achromatic). Whilst Eqn 1 describes how the retardation is an inverse function of wavelength, the effective birefringence of the R8 cell increases at longer wavelengths, cancelling out the effect of the change in wavelength and creating a constant retardation (Roberts et al., 2009). This wavelength-independent effective birefringence of the R8 cell occurs through a combination of the intrinsic birefringence of the microvillar membranes and a form birefringent component owing to the ordered structure of subwavelength-sized components (Born and Wolf, 1999). The effectiveness is therefore controlled by measureable characteristics of the cell: the length of the R8 rhabdom, the diameter of the microvillar tubes and the volume packing fraction of the microvilli in the cell (Fig. 2B) (Marshall et al., 1991). Changes to either property alter the overall effective birefringence of the cell.
The reliance on cell size for a specific retardance function is itself intriguing. When comparing different species of stomatopod, differences in body size exist: adult Haptosquilla trispinosa are usually less than 35 mm in length compared with adult Lysiosquillina maculata, which can grow to more than 30 cm. Large size differences also exist developmentally within a species. For example, the post-larvae of L. maculata are smaller by more than a factor of 10 compared with the adults. There are concomitant differences in eye size and internal eye anatomy, and changes to eye size could limit the space available to the R8 cell both optically and anatomically (Marshall et al., 1991). Any such changes may affect the retardance and optical function of the cell, unless one factor is adjusted relative to another, and investigating these changes was a second aim of this work. We also set out to quantify the potential differences in R8 cell size across different species of stomatopod and determine what effect body size has on the cells' function as a quarter-wave retarder. Attention was paid to differences that may occur within species, such as differences between males and females, variation in size between individuals, and finally any variation within eyes according to photoreceptor placement and packing.
MATERIALS AND METHODS
Stomatopods were collected from coral reef and rubble, and mangrove areas around Lizard Island (14°40′40.8″S, 145°26′48.1″E) at a depth range of 0.5–5 m [Great Barrier Reef Marine Park Authority (GBRMPA) permit no. G12/35005.1, Fisheries Act no. 140763]. Species collected were Gonodactylaceus falcatus, Gonodactylus smithii, Haptosquilla trispinosa and Lysiosquillina maculata. Two further species, Odontodactylus scyllarus and O. latirostris, were collected from Shag Rock off North Stradbroke Island (27°25′0″S, 153°32′59.9″E) at a depth range of 10–20 m (Moreton Bay Marine Park permit no. QS2013/CVL625). The animals represent species of a wide body length range (20–200 mm) in the Gonodactyloidea and Lysiosquilloidea superfamilies. Animals were either anesthetized and dissected at the field site, or transported live to the holding aquaria at the University of Queensland, St Lucia.
Retinal cell measurements
The animals were anesthetized by cooling on ice, decapitated, and both eyes were removed and fixed in 4% paraformaldehyde and 2.5% glutaraldehyde (in PEMS buffer) (adapted from Chiou et al., 2005). Samples were post-fixed in 1% osmium tetroxide, and dehydrated in solutions with increasing ethanol concentrations. Samples were then infiltrated and embedded in EPON resin blocks for transmission electron microscopy (TEM). A microwave regime (Pelco Biowave, Ted Pella, USA) was used to aid dehydration (1 min at 150 W for each ethanol concentration) and infiltration (3 min at 150 W under vacuum).
Embedded eyes were sectioned using an ultramicrotome (EM U26, Leica, Germany). Ultrathin (∼60 nm) and semithin (∼500 nm) coronal sections were obtained through the entire rhabdom in rows 5 and 6 of the midband (Fig. 2). Ultrathin sections were collected on copper grids and stained with 5% uranyl acetate in 50% ethanol and Reynolds lead citrate for viewing in TEM (JEM-1010, Jeol, Japan). Semithin sections were collected on glass slides and stained with Toluidine Blue for light microscopy (Bx61, Olympus, Japan).
Images of R8 cells were collected from the high acuity region of the eye (Marshall et al., 1991), along with other regions for comparison, using both TEM and light microscopy, which allowed for measurements to be made for both the length of the R8 cell and the width of the microvilli. Light microscopy was also used to make a separate set of measurements of the length for both the R8 and the R1–7 rhabdom. The total length of the rhabdom was obtained by adding the R8 and R1–7 measurements. Measurements were acquired from TEM images using iTEM imaging software (EMSIS, Muenster, Germany). To record the length of R8 cells across the whole eye, serial sections were obtained for light microscopy through the entirety of rows 5 and 6. Images were registered in Fiji (Schindelin et al., 2012) and the positions of R8 cells were digitized using custom MATLAB-based scripts (MathWorks, Natick, MA, USA).
Model development and calculation of R8 retardation values
Using the methods developed by Roberts (2006) and Roberts et al. (2009), two-dimensional surface plots of retardation values were calculated based on R8 cell lengths and R8 microvilli diameter along with assumed refractive indices (Roberts et al., 2009). The surface plots allow direct visualization of the parameter sets that combine to create a specific retardation value. The contour lines depict particular values of retardance, with the bold line indicating the measurements required for quarter-wave functionality (0.25). All calculations were performed using R (version 3.2.5) and the script is available upon request. Effective dielectric tensors (Bêche and Gaviot, 2003) were calculated for microvillar diameters in the range 20–85 nm and photoreceptor lengths 40–120 μm. Onto these plots were mapped the experimentally measured values.
Data were analysed using a standardized major axis regression in R (version 3.2.5), using the smatr package, to determine the relationship between the entire rhabdom length and the length of the R1–7 and R8 rhabdoms across and between species.
Total rhabdom measurements
Measurements of the R1–7 and R8 rhabdoms were completed using 16 individuals from five species of stomatopod: Gonodactylus smithii (n=5), Odontodactylus scyllarus (n=2), O. latirostris (n=3), Lysiosquillina maculata (n=3) and Haptosquilla trispinosa (n=3) (Table 1, Fig. 3). Measurements of the length of the R1–7 rhabdom and the R8 rhabdom were obtained from row 5 in all individuals and in row 6 for 13 individuals. An average total rhabdom length was calculated for each individual by averaging all measurements from rows 5 and 6. For each individual between three and six measurements were made for each row.
Conservation of R8 cell length
The rhabdom length varies between individuals based on body size (Fig. 3). Larger individuals have longer rhabdom lengths than smaller individuals. R1–7 cells account for the majority of the rhabdom and the length can vary greatly both within and between species. The R8 cell, however, is more consistent in length, with only small variability between individuals of differing body lengths (Fig. 3). Plotting the total rhabdom length against the length of the R1–7 cells shows a strong correlation between variables (r=0.99), which remains strong when considering individual species (Fig. 4A). The relationship between the total rhabdom and the R8 length is weaker (r=0.54), indicating that the R8 cells are more constrained in length than the R1–7 cells (Fig. 4B). Although the slopes for individual species do vary, the relationship between the R8 cell length and the total rhabdom length remains weak in all species.
R8 measurements for use in the calculation of the retardation
To further investigate the relationship between the length of the R8 rhabdom and the width of the individual microvilli within each R8 cell (Fig. 2A,B), a second set of measurements were made from 25 individuals from six species of stomatopod: G. smithii (n=7), G. falcatus (n=3), O. scyllarus (n=3), O. latirostris (n=3), L. maculata (n=4) and H. trispinosa (n=5) (Table 2). These measurements were obtained using only light microscopy preparations where the whole length of the rhabdom could be measured. Measurements were obtained from 24 of the included individuals for row 5, and from 22 for row 6. Measurements are only included here if they were obtained for both the length of the R8 and the width of the microvilli in a single row. The number of measurements made for each individual varied (from two to 15 measurements for R8 length, and 14 to 45 measurements for microvilli width) based on the alignment of the sections through the R8. The average R8 length and microvilli width were calculated for each individual (Table 2).
Variation between species
R8 cells vary slightly in their birefringent properties between species, although for most, this has little effect on their function as quarter-wave retarders. Fig. 5 illustrates the birefringent properties of the R8 cells in each species. Odontodactylus scyllarus, O. latirostris, L. maculata, G. smithii and G. falcatus (Fig. 5) all have R8 cells that fall close to the line that the model predicts will provide quarter-wave retardance, resulting in a circular polarization sensitivity in rows 5 and 6. In H. trispinosa, the R8 cell measurements fall closer to the 0.2 line, suggesting that these cells convert CPL into EPL.
Variation within species
There was some variation in the size of the R8 cells, in relation to the length of the animal. This mostly affected species that displayed a large variation in body length, such as in L. maculata (which varied in body length from 32 to 200 mm) and in O. scyllarus (which varied from 45 to 167 mm). The corresponding R8 cells varied from 64.03 to 85.80 µm in L. maculata and from 41.87 to 95.54 µm in O. scyllarus, and resulted in variation in the cells’ calculated birefringence (Fig. 5). Essentially, small, presumably juvenile animals in these species possess R8 cells that would not convert circular polarization to linear but to EPL, resulting in a weaker stimulation of the underlying R1–7 cells. In the case of the very small 45 mm O. scyllarus, this would be around half that of a perfect quarter-wave retarder.
The smaller species, such as G. falcatus, G. smithii, H. trispinosa and O. latirostris, show less variation in R8 cell measurements, while also having less variation in body length, at least for the individuals we caught. It remains possible that very small post-larval individuals in these species might display similar birefringence variation. For G. falcatus, G. smithii and O. latirostris, all the measurements within each species fall near the 0.25 line, indicating that they will be able to function as quarter-wave retarders. Haptosquilla trispinosa R8 cells fall close to the 0.2 line, suggesting either an elliptical sensitivity converting to LPL or just less efficient sensitivity to CPL, as suggested for the smaller individuals of the larger species above.
Variation within the retina
The properties of the R8 can also vary across the length of the midband in a single eye. To investigate the variation within the eye, measurements across the length of the midband were made in two speceis, G. smithii and H. trispinosa. The length of the R8 cells in rows 5 and 6 in G. smithii varied across the eye, with the cells being significantly longer in the inner part of the eye compared with the outer part (Fig. 6A). Note, however, the larger variation in R1–7 cells over this range. In H. trispinosa, however, the R8 cell measurements remained a more constant size in the inner and outer part of the eye (Fig. 6B). When comparing these results to the birefringent model using the microvilli width coupled with the length of the R8 cells (Fig. 7), it appears that while the R8 cells in H. trispinosa function similarly across the whole midband, the R8 cells in G. smithii vary greatly in their ability to function as quarter-wave retarders across the midband. Although the inner cells are longer than the optimal length, the outer cells are too short, and it is cells in and around the acute zone that have R8 cells that fall within the correct size range for good quarter-wave retardance.
The results of this study illustrate that, while there are several sources of variability in the size and thus the birefringence of the R8 cells across stomatopod species, R8 cells appear more conserved than other cells in the retina. This suggests adaptive pressure for R8 cells to stay within a certain size range to maintain the ability to act as quarter-wave retarders. There is a strong relationship between the length of the R1–7 cells and total rhabdom length, but less of a correlation between the R8 length and the total rhabdom length. This relationship highlights that the R8 length is conserved when the total size of the rhabdom increases.
Despite investigating several species of stomatopods with a large range of body sizes, habitat preferences and ecological constraints, the size of the R8 cells remained largely similar. As noted, this suggests significant adaptive pressure to keep these cell types within a functional size range for quarter-wave retardance. The exception to this is H. trispinosa, where the R8 cells’ retardance suggests sensitivity to an elliptical form of polarized light with retardance closer to 0.2 than 0.25. Why H. trispinosa would be tuned to detect elliptical polarized light (EPL) is not yet known, but it is fascinating to speculate that its eye may be tuned more specifically to a particular ellipticity in the environment. Alternatively, selection for precise quarter-wave retardance in this species may be relaxed, as CPL is less significant for its current survival needs. Notably, this species does display strong linear polarization signals (Cronin et al., 2009).
Within a species, most cells perform well as quarter-wave retarders, but there was some individual variability, particularly where the total body size of the animals varied greatly from juvenile to adult, such as L. maculata and O. scyllarus. This indicates that, while large individuals of a species detect CPL, small individuals of the same species, perhaps not yet sexually mature and with no need to pay attention to circular polarization, may not. In these cases, instead of opting for a shorter R1–7 rhabdom, the birefringent function of the R8 cell is compromised, suggesting that a critical R1–7 length is required for the rhabdom to maintain sensitivity. Additionally, once the R8 has reached the point of quarter-wave retardance it no longer continues to increase in size, i.e. the R8 seems to stop increasing in size at sensitivity to circularly polarized light, allowing large relative increases in the R1–7 rhabdom.
Sensitivity changes within the eye
Investigation of R8 sizes across the midband in H. trispinosa and G. smithii provided evidence that R8 cell length can also vary within an individual eye. While the length of the R8 cells in H. trispinosa displayed a similar size across the midband, G. smithii displays more variability in R8 cell length, with the inner, more forward-facing part of the midband having longer R8 cells than the outer, lateral part. In fact, G. smithii only have circular polarization sensitivity in the central part of the midband, in ommatidia around the acute zone. It may be that the elongate design of this eye places spatial limits on the ommatidial dimensions and that retaining good CPL sensitivity in the forward-facing zones is enough. The eye of H. trispinosa is evenly spherical along the midband direction (as are the eyes of all other species examined; Marshall et al., 1991; Marshall and Land, 1993), making it possible to retain the same R8 dimensions along the whole midband, including the acute zone, without compromising the size of the R1–7.
Function of circular polarization vision
Previous behavioural tests have shown that O. scyllarus can learn left from right circular polarization in feeding trials (Chiou et al., 2008). It is also known that G. falcatus prefers not to enter burrows emitting CPL (Gagnon et al., 2015), although the actual use of CPL in either of these species or others included in this study is still hypothetical. It is not known, for instance, if food items of O. scyllarus reflect CPL differentially, or if this species just manages to transfer its ability for discrimination to an isolated feeding circumstance in behavioural trials. Odontodactylus species and specifically G. falcatus are known to reflect CPL (Chiou et al., 2008; Gagnon et al., 2015) and one possible function is inter- or intraspecific signalling. However, this is yet to be directly tested. In the context of other animals, scarab beetles have been known for many years to reflect CPL (Michelson, 1911) and initial behaviour evidence suggested sensitivity to CPL in American jewel beetles (Brady and Cummings, 2010). However, this sensitivity was not found in a more recently conducted survey of four related European species (Blahó et al., 2012). If CP is used by stomatopods for inter- or intraspecific signalling, it should be noted that so far only Odontodactylus cultrifer, G. falcatus, Squilla mantis and some Neogonodactylus species are known to exhibit strongly polarized CP reflections (Chiou et al., 2008; Gagnon et al., 2015), but further investigation is underway to quantify polarized signals more thoroughly in other species.
LPL-insensitive detection channels
Finally, it is worth considering that the R8 cells in rows 5 and 6 of the midband may not have originally evolved with the primary purpose of converting CPL to LPL. The R8 cells in rows 5 and 6 are sensitive to ultraviolet wavelengths of light because the R8 of row 5 is orientated perpendicular to the R8 of row 6, and by comparison they act as an ultraviolet PS channel (Kleinlogel and Marshall, 2009) in the midband sensitive to LPL. However, if the R1–7 photoreceptors below are primarily concerned with colour vision or a measure of intensity (Marshall et al., 1991; Thoen et al., 2014), then elimination of PS is ideal and allows them to potentially be subject to false colours (Kelber et al., 2001) or false signals. Other arthropods avoid sensitivity to LPL by (1) producing rhabdomeres with microvilli that are misaligned, (2) forming twisted rhabdomeres with microvilli of low birefringence or (3) summing (in the lamina) the signals of receptors with mutually perpendicular microvilli (Eguchi and Waterman, 1967, 1973; Waterman, 1981; Marshall et al., 1991). The use of the R8 cell as a quarter-wave retarder provides a novel approach to remove the linear polarization information in the signal. Interestingly, this way of removing unwanted LPL is also used in photography, where CPL rather than LPL filters are used (Goldberg, 1992). A clear, although again hypothetical, evolutionary progression could therefore drive a quarter-wave retarder to be exploited through the emergence of CPL signals to provide certain species with the sensory ability to detect CPL.
While several alternative, and not necessarily mutually exclusive, hypotheses exist for the evolution of this apparently complex polarization vision system, it is worth re-iterating that our knowledge of what the stomatopods actually do with polarization vision and polarized reflections, where they exist, is still in its infancy. Field observations and behavioural analyses would help place such hypotheses on a firm foundation of reality.
The authors wish to acknowledge the facilities, and the scientific and technical assistance, of the Australian Microscopy and Micro analysis Research Facility at the Centre for Microscopy and Microanalysis, The University of Queensland, and Dr Hanne Thoen for her comments on the manuscript.
Conceptualization: R.M.T., M.J.H., N.W.R., T.-H.C., J.M.; Methodology: R.M.T., M.J.H., N.W.R., T.-H.C.; Software: N.W.R.; Formal analysis: R.M.T., M.J.H., N.W.R.; Investigation: R.M.T., M.J.H.; Writing - original draft: R.M.T., N.W.R.; Writing - review & editing: R.M.T., M.J.H., N.W.R., T.-H.C., J.M.; Supervision: J.M.; Funding acquisition: N.W.R., J.M.
This work was supported by grants awarded to J.M. by the Asian Office of Aerospace Research and Development (AOARD-12-4063) and the Australian Research Council (FL140100197), and to N.W.R. by Air Force Office of Scientific Research (AFOSR FA8655-12-2112).
The authors declare no competing or financial interests.