Out of the blue: the evolution of horizontally polarized signals in Haptosquilla (Crustacea, Stomatopoda, Protosquillidae)

Polarization sensitivity is a common visual specialization that has evolved in both terrestrial and aquatic animals, and is particularly prevalent in invertebrates (Wehner and Labhart, 2006). On land, many insects use the celestial polarization pattern for navigation (Wehner, 1976; Rossel and Wehner, 1986; Labhart and Meyer, 1999; Dacke et al., 2003), while in the ocean, some crustaceans and cephalopod molluscs use polarization information to detect prey and possibly as a means of conspecific communication (Shashar et al., 1996; Cronin et al., 2003a; Chiou et al., 2007; Mäthger et al., 2009; Cronin et al., 2009; Chiou et al., 2011). In the context of communication, polarization often forms composite signals with other visual dimensions, such as hue and brightness (Cronin et al., 2003a; Cronin et al., 2009).

The term polarization is used to define several properties of light. The angle of polarization describes the predominant direction in which the electric field of the light oscillates, while the degree of polarization defines the extent to which waves oscillate at the same angle. Underwater, differential sensitivity to either angle or degree of polarization has several fundamental advantages over other forms of visual information (Cronin et al., 2003a; Cronin et al., 2003b; Cronin et al., 2009; Shashar et al., 2011). For instance, in shallow, clear marine waters, the intensity and spectral composition of the downwelling light can vary dramatically, both as a function of the time of day, and because of environmental factors such as turbidity (Cronin et al., 2014). In such changing conditions, the polarization of light remains more constant than other visual dimensions over short ranges (Waterman, 1954; Cronin, 2001), which renders it a reliable provider of information (Shashar et al., 2011; Johnsen et al., 2011). Previous research in this field has focused on either the underlying retinal mechanisms of polarization sensitivity (for reviews, see Horváth and Varjú, 2004; Roberts et al., 2011) or the optical mechanisms by which polarization and multi-component polarization and/or colour signals are produced (Chiou et al., 2005; Mäthger and Hanlon, 2006; Chiou et al., 2007; Mäthger et al., 2009; Cronin et al., 2009). In contrast, the evolutionary context of polarization signal content relative to the visual system of receivers is still very much unknown.

Stomatopod crustaceans are some of the best-studied species in terms of polarization vision. Electrophysiological studies have detailed the spatial variation of polarization sensitivity in the different photoreceptor classes in the eye (Kleinlogel and Marshall, 2006; Chiou et al., 2008). Optical measurements (Marshall et al., 1991; Chiou et al., 2008), optical modeling (Roberts et al., 2009) and molecular methods (Porter et al., 2009; Roberts et al., 2011) have provided additional information on the underlying mechanisms of polarization sensitivity. Optical techniques have also shown that many species of stomatopod produce visual signals that are either linearly or circularly polarized (Chiou et al., 2005; Chiou et al., 2008; Cronin et al., 2009). The stomatopod genus Haptosquilla (family Protosquillidae) is known to use signals from the first maxillipeds for both sexual and agonistic communication (Dingle and Caldwell, 1969; Caldwell and Dingle, 1975; Chiou et al., 2011). A common feature of Haptosquilla first maxillipeds is the production of a conspicuous blue structural reflection (Chiou et al., 2005; Cronin et al., 2009). Fig. 1 illustrates the blue signal in four species: Haptosquilla trispinosa, H. glyptocercus, H. stoliura and H. banggai. In some species of the genus (e.g. H. trispinosa, H. stoliura and H. banggai), this reflection is also horizontally polarized (Chiou et al., 2005; Cronin et al., 2009).

Here we explore the potential evolutionary pathways of polarization communication in protosquillid stomatopods. First, we use experiments to investigate whether the behavioural responses to different forms of polarization signal content are species specific. We do this by exploiting the animal's innate behavioural responses to polarized looming stimuli presented on modified LCD monitors. We compare four representative protosquillid species: H. trispinosa (Dana 1852), H. glyptocercus (Wood-Mason 1875), Chorisquilla tweediei (Serène 1950) and C. hystrix (Nobili 1899). Second, and in the context of the signal's polarization content, we measure the threshold at which H. trispinosa are no longer able to discriminate between two different angles of polarization. Finally, we construct a phylogeny of protosquillid species to consider the evolution of the polarization properties of maxilliped signals.

Haptosquilla trispinosa, H. glyptocercus and C. tweediei all showed a significantly greater probability of response to the horizontally polarized stimulus compared with a vertically polarized stimulus (H. trispinosa: Wilcoxon test: Z=2.93, d.f.=9, P=0.002; H. glyptocercus: Z=2.42, d.f.=9, P=0.02; C. tweediei: Z=2.77, d.f.=9, P=0.004; Fig. 2A–C). Chorisquilla hystrix also appeared to be more responsive to horizontally polarized light (Fig. 2D), but the small sample size (n=5) precluded statistical testing. There was no significant difference between H. trispinosa, H. glyptocercus and C. tweediei in their relative responses to the two stimuli (Kruskal–Wallis test: χ²=2.90, d.f.=2, P=0.24).

Haptosquilla trispinosa showed little or no response to stimuli when the difference between the polarization angles of the stimulus and background was between 31.4 and 20 deg (Fig. 3; supplementary material Table S1). At angles of 20 deg or less, the animals rarely responded to the polarization stimulus; at values of 31.4 deg and above, they displayed a consistent statistically significant response to the stimulus.

The first maxilliped reflections from H. trispinosa, H. glyptocercus, C. tweediei and C. hystrix are presented in the microscope images displayed in Fig. 4. Both H. trispinosa (Fig. 4A) and H. glyptocercus (Fig. 4B) show blue reflections from the maxillipeds compared with very weak, spectrally broad reflections from the Chorisquilla species (Fig. 4C,D). Of the blue Haptosquilla reflections, H. trispinosa are horizontally polarized (Fig. 4A) whereas the reflections from H. glyptocercus are unpolarized (Fig. 4B).

Visual analyses of other species of Haptosquilla showed that H. stoliura, H. banggai, H. pulchella, H. nefanda and H. hamifera all have blue-reflecting first maxillipeds, but only the reflections from H. stoliura, H. banggai, H. pulchella and H. nefanda are horizontally polarized. Within the rest of the Protosquillidae, five further species have been analyzed (C. excavata, C. hystrix, C. tweediei, Echinosquilla guerinii and Protosquilla folini), with none possessing blue or blue and horizontally polarized first maxillipeds. Outside of the Protosquillidae, six other stomatopod species from nine genera and four families have been inspected for first maxilliped signal types. Of these species, only G. smithii possess blue signals and no other species possess either blue or horizontally polarizing signals (Fig. 5).

Phylogenetic analyses of protosquillid relationships recapitulate previous studies (Barber and Boyce 2006; Porter et al., 2010) recovering the protosquillids [bootstrap percentages (BP)=98], and in particular the genus Haptosquilla (BP=89), as monophyletic (Fig. 5). Within the Haptosquilla, our phylogeny recovered two sub-groups of species that correspond to the two known types of first maxilliped signaling, either blue and unpolarized or blue and polarizing.

Our results provide direct evidence that several species of stomatopod have an inherent (i.e. non-trained) behavioural response to a looming, linearly polarized stimulus. Moreover, all the protosquillid species tested displayed a greater probability of response to horizontally polarized stimuli compared with those that are vertically polarized. The measurements of the structural colour and polarization properties of the maxillipeds, in combination with the comparative phylogenetic analyses, revealed that of these protosquillids, only the genus Haptosquilla displays the blue signals. Furthermore, it is only the sub-group of Haptosquilla including H. trispinosa that possesses the additional polarized signal dimension. In these species, the polarization of the signals is always orientated horizontally. Therefore, it is possible that the common behavioural predisposition towards horizontally polarized stimuli seen across the protosquillids could have biased the polarization content of first maxilliped signals to be horizontal in the H. trispinosa clade (Guilford and Dawkins, 1991; Endler and Basolo, 1998). A common question raised by the concept of sensory bias is why does the bias pre-exist? Whilst we can only speculate, the bias for a horizontal angle of polarization may come from the fact that this angle is most prevalent in reflections from objects and preferential sensitivity may have previously evolved to improve contrast discrimination (Temple, 2012).

Haptosquilla trispinosa also displayed a threshold of between 21.4 and 30 deg in their response to distinguishing between two angles of polarization. Such a coarse level of discrimination would improve the signal-to-noise ratio of a linearly polarized signal by effectively low-pass filtering any variation in the background. This threshold is an order of magnitude higher than measured in other species [fiddler crab, Uca vomeris, 3.2 deg (How et al., 2012); cuttlefish, Sepia plangon, 1 deg (Temple et al., 2012)] and is suggestive of tuning for high contrast signals compared with the current evidence that other crustacean and cephalopod polarization visual systems are used to resolve high levels of polarization detail.

The complex nature of stomatopod eye design (two hemispheres separated by a specialized midband) may place limitations on the amount of information that can be processed from the visual scene but in turn enhance the processing efficiency. Currently, it is thought that the two hemispheres are primarily involved in producing a two-dimensional representation of the visual scene, over which the midband is then scanned, rather like a line-scan sensor, to expand on the colour and polarization information (Land et al., 1990). The motion component of the LCD looming stimulus used in our experiment is therefore most likely to be stimulating responses in the stomatopod visual hemispheres, which elicit a visual saccade to the target, and presumably this would be followed by a subsequent visual scan of the target with the midband to fill in the remaining information. Therefore, it is conceivable that much of the early visual information is simplified to speed up sensory processing [for an equivalent discussion for colour vision, see Thoen et al. (Thoen et al., 2014)]. If so, the polarization discrimination responses we have measured specifically represent a property of the visual system in the dorsal and ventral hemispheres. However, the precise behavioural context should also not be ignored. It is quite possible that the measured discrimination threshold is specific to the task demanded of the animals. Further work is also still needed to investigate how the degree of polarization affects behavioural responses to such polarization signals.

Overall, our findings provide a framework for understanding the potential evolutionary pathway of the polarization properties of these maxilliped signals in stomatopods. Successful communication relies on information being sent through the environment in such a way that it will be received in its intended form, and be interpreted as to elicit a behavioural response in the intended receiver (Parten and Marler, 2005). In this context, the selective pressures on signal evolution are both efficacy-driven and content-driven (Guilford and Dawkins, 1991; Hebets and Papaj, 2005). As described in the Introduction, polarization provides a reliable form of visual information, particularly in spectrally variable light environments, such as those in which these species of stomatopod reside. The increase in signal efficacy by the inclusion of this extra visual dimension is therefore fairly clear. The behavioural bias towards horizontally polarized light provides a further explanation for why the polarized content of the signals has evolved to be horizontally polarized. Together, the addition of polarization to the signal and the nature of the bias suggest both the proximate and ultimate driving principles, respectively, for the evolution of this complex signal.

Two questions for the future are: (1) can manipulating the relative polarization contrast of the signal and the background influence the bias; and (2) do the spectral and polarization dimensions act independently for purposes of information redundancy or do they combine in a functional way, for example, increasing the accuracy of receiver response as is described by an amplifier hypothesis of multi-component signals (Hasson, 1991; Candolin, 2003; Hebets and Papaj, 2005)? We suggest that future studies of combined polarization and colour signals in other animals should also carefully consider how these dual dimensions are viewed together by receiver visual systems under the correct environmental light conditions. Whilst it is not always easy to decompose complex signals and test the functions of individual components (Hebets and Papaj, 2005), the combined colour and polarization signals in stomatopods represent an excellent behavioural system to investigate the function and evolution of signal complexity.

To investigate the inherent ability of stomatopods to generate distinct behavioral responses to polarized stimuli, we collected 39 individuals of H. trispinosa, 10 individuals of both H. glyptocercus and C. tweediei, and five individuals of C. hystrix from offshore reefs near Lizard Island, Great Barrier Reef, Australia, in August 2011 (Queensland–GBRMPA permit G12/35042.1). Animals were maintained before testing in a natural seawater flow-through marine aquarium facility at the Lizard Island Research Station (24–25°C, natural daylight illumination, and fed pieces of frozen shrimp). All procedures were approved by the Animal Ethics Committees of the University of Queensland [AEC, permit no. QBI/223/10/ARC/US AIRFORCE (NF)].

Individual stomatopods were placed in a 30×15×15 cm tank containing local beach sand. Each individual was placed inside an 8 mm diameter clear tube and restrained using a small amount of fishing line (Land et al., 1990; Cronin et al., 1991). The animal was positioned such that the eyes were forward of the front end of the tube (Fig. 6A). Directly above the animal was a video camera (Canon Legria FS20) that recorded its response to the presentation of the stimuli. On the outside of the tank, and in front of the animal, was an LCD screen (Viglen LC552; 1280×1024 pixel spatial resolution at 60 Hz); the eyes were at a distance of ~12 cm from the screen. By removing the front polarizer from the LCD screen and addressing the LCD with a grayscale value of either 0 (black) or 255 (white), the local output polarization could be controlled as vertical (V stimulus) or horizontal (H stimulus), respectively (Pignatelli et al., 2011). The stimuli expanded to cover 22.5 deg of the visual field angle in 1 s (taking into account refraction at the air/glass/water boundaries). The simple electro-optic control of the polarization of the light permitted not only dynamic control of the polarization, but most importantly an inherent zero luminance and chromatic contrast between the background and the looming stimulus. To check the polarization properties of the LCD, accurate broadband Stokes parameter measurements (Fig. 6B) were made using Glan–Thompson polarizers and a ¼ wave Fresnel rhomb (Edmund Optics, York, UK), which permitted the computation of the polarization ellipse of each of the stimuli for any wavelength (Fig. 6C).

All animals received a balanced pseudo-randomized presentation of 10 H stimuli and 10 V stimuli, against a perpendicularly linearly polarized background. No more than three instances of the same stimulus were presented in a row. We randomly varied the time between successive stimuli, from 20 to 120 s, to minimize any effect of habituation. To determine whether the animal responded to the two stimulus types, we monitored the optokinetic response of the focal animal. We defined a positive response to the stimulus as a saccadic eye movement, in which one or both eyestalks were rapidly brought together (see Fig. 7 for an example). No such saccadic eye movements were observed in a 5 s period before the onset of the stimulus or from 3 s after its presentation. Animals were scored by their number of responses out of the 10 presentations, giving a probability of response to each stimulus type.

A similar method was used to measure the polarization angular contrast sensitivity of H. trispinosa. Individual unrestrained animals were housed in a 20×20×30 cm aquarium partition in burrows positioned ~12 cm from the front wall. A different polarization LCD monitor [HP L1906; see How et al. (How et al., 2012) for calibration details] to that described above, but with very similar properties, was positioned against the front wall. A looming circle stimulus expanded to cover 27 deg of the visual field angle in 1 s (taking into account refraction at the air/glass/water boundaries). The greyscale values addressed to the monitor were set to 0 (black) for the background and ranged between 0 and 255 for the stimulus, resulting in a stimulus that varied in the angle of polarization against a horizontally polarized background, with no corresponding changes in hue or light intensity. Stomatopod eye movements in response to the stimulus were recorded using a digital video camera (Sony HDR-SR11, Tokyo, Japan) mounted on the top edge of the front aquarium wall. Stimuli were generated automatically using MATLAB (r2011, MathWorks, Natick, MA, USA) and the whole experiment was conducted without experimenter intervention. Video recordings were synchronized to the stimulus by means of an audio signal conveyed by audio cable directly from the computer to the microphone port of the camera. Measures of saccadic eye movements were made in a 5 s period both before and after the stimulus presentation. Two independent groups (n=15 and 14 animals) were tested using two sets of stimuli (angles of 0, 0.5, 1, 5, 7, 9 and 11 deg and 20, 31, 56, 70 and 74 deg, respectively). The stimulus order was fully randomized and the interval between stimuli was randomized between 20 and 60 s.

Images of the maxillipeds of H. trispinosa, H. glyptocercus, C. tweediei and C. hystrix were taken though a Leitz compound microscope (Leica Microsystems, Wetzler, Germany) using a 10× objective and a Canon G9 digital camera (Canon, Melville, NY, USA) mounted using a photo tube extension on the trinocular head. Spectral reflection data of the same four species were measured using an Ocean Optics halogen HL-2000 light source (Ocean Optics, Dunedin, FL, USA) mounted at the back focal plane of the eyepiece and illuminating the maxillipeds normally. The reflected light was collected at the back focal plane of the second eyepiece using a 1 mm diameter optic fibre connected to a QE65000 spectrometer (Ocean Optics). Linear horizontal and vertical polarization filters were placed in the path of the reflected light inside the microscope to collect each respective polarized reflectance spectrum. Over several preceding years, the colour and polarizing nature of the first maxillipeds from 17 other representative species of stomatopods across the superfamily Gonodactyloidea have been assessed visually by viewing the maxillipeds thorough a rotatable linear polarizer.

To investigate the potential evolutionary pathway of color and polarization signals within the genus Haptosquilla, DNA sequences from both nuclear and mitochondrial genes for all available species were obtained from GenBank, provided by P. Barber (Barber and Boyce, 2006) or were sequenced following the methods of Porter et al. (Porter et al., 2010) (supplementary material Table S2). Additional representative stomatopod species from within the same family (Protosquillidae) and superfamily (Gonodactyloidea) were included to provide increased resolution and stability at deeper nodes within the phylogeny and to use as outgroups. We used a concatenated matrix consisting of nucleotide sequences from the cytochrome oxidase I (COI) and 16S mitochondrial genes, and the 18S and 28S nuclear rDNA genes, although the number of sequences available varied across species (see supplementary material Table S2 for full description of data sources and gene representation).

Nucleotide sequences of the 16S, 18S and 28S genes were aligned using the E-INS-I strategy in MAFFT v6.0.0 (http://mafft.cbrc.jp/alignment/server/) (Katoh et al., 2002; Katoh et al., 2005). The COI sequences were inspected for evidence of pseudogenes (e.g. stop codons, indels not continuous with codons) and then manually aligned using the translated amino acid sequences. The four gene regions were then concatenated and the combined dataset was used to reconstruct a phylogeny using Randomized Axelerated Maximum Likelihood (RAxML) v.7.2.7 with rapid bootstrapping as implemented on the Cyberinfrastructure for Phylogenetic Research (CIPRES) Portal v.2.0 (Stamatakis 2006; Stamatakis et al., 2008; Miller et al., 2009). Three partitions were designated for the RAxML analysis: (1) COI codon positions 1 and 2; (2) COI codon position 3; and (3) all of the ribosomal genes (16S, 18S and 28S). All partitions were analyzed with the GTR+gamma model, as this was the best-fitting model available in RAxML, according to the results of jModelTest v0.1.1 (Guindon and Gascuel 2003; Posada 2008).

All statistical analyses were conducted in R 3.0.2 (R Foundation for Statistical Computing). Response probabilities to either horizontally or vertically polarized looming stimuli were analysed using Wilcoxon signed-rank tests and differences between species were calculated using a Kruskal–Wallis rank sum test. The individual saccadic responses of H. trispinosa to different angular e-vector contrasts were analysed using a McNemar's test.

We are very grateful to the Lizard Island Research station staff for all their assistance, to Innes Cuthill for help with the statistical analysis, and to Paul Barber for providing sequence data.