Dark-eyed females: sexually dimorphic prespawning coloration results from sex-specific physiological response to hormone exposure in the sand goby Pomatoschistus minutus (Gobiiformes: Gobiidae)

Abstract

The function and regulation of female nuptial colour signals are poorly understood. In fish, colour is often mediated by chromatophores, allowing for rapid and versatile signalling. Here, we examine a distinct but temporary black line around the eyes and snout (‘dark eyes’) displayed by female sand gobies before spawning and never observed in males. We investigate the regulatory mechanism of the display by analysing the number of melanophores in both sexes in vitro and their response to hormonal exposure. We also test the hypothesis that dark eyes serve an anti-glare function and focus the line of sight, by analysing the frequency, intensity and duration of the display in bright and dim light, with and without males present. We show that the sexes do not differ in terms of the number of melanophores, but that males and females respond in different ways to exposure to melanocyte-stimulating hormone, which has a stronger dilatory effect in females and results in a darker line. However, the darkness of the iris is not affected. Neither light levels nor the presence of potential mates affect the frequency of the dark eye display, but the display is longer lasting and more intense in the presence of smaller nest-holding males.

INTRODUCTION

Coloration in animals can range from the spectacularly flamboyant to the superbly cryptic and can serve a variety of functions, such as camouflage and predator diversion or deterrence, aposematism and mimicry, in addition to social signalling to indicate aggression, social status or reproductive state (Andersson, 1994; Höglund et al., 2002; Lin et al., 2009; Nilsson Sköld et al., 2016). The most extravagant body colours are often present during reproduction, as nuptial coloration. In general, exaggerated nuptial coloration is most prominent in males, and owing to positive relationships between display traits and male mating success, and hence fitness, it is typically ascribed to sexually selected ornaments, i.e. traits that serve to attract mates but lack a function outside these interactions (Andersson, 1994; Tobias et al., 2012; Schlupp, 2018). Nonetheless, display traits can also serve a function in intrasexual competition beyond attractiveness, in which case they are referred to as status badges (Andersson, 1994; Tobias et al., 2012). Duller female nuptial coloration can then be viewed as a side-effect of male ornamental coloration by way of conspecific genetic similarity (Lande, 1980, 1987; Kraaijeveld et al., 2007). A similar reasoning is applied to explain conspicuous female nuptial coloration in species where sex roles are either reversed or dynamic, i.e. where females compete for mates (Berglund, 1991; Amundsen & Forsgren, 2001; Kraaijeveld et al., 2007; Edward & Chapman, 2011; Muck & Goymann, 2011). Although females can benefit from polyandry (Arnqvist & Nilsson, 2000; Hosken & Stockley, 2003; Taylor et al., 2014), the relationship between fitness and the number of mates is likely to be more complicated in females compared with males. Still, a growing body of work has shown that conspicuous female nuptial coloration can enhance female fitness, e.g. by signalling status in intrasexual interactions, in competition over key resources or in defence of territory (Heinsohn et al., 2005; LeBas, 2006; Rosvall, 2011; Tobias et al., 2012).

Bright colours and sexually dimorphic traits are not necessarily ornamental or involved in intrasexual interactions, but can evolve in response to other selection pressures that might differ between the sexes (Hedrick et al., 1989; Schlupp, 2018). For example, disruptive coloration can provide camouflage and eye-spots can help to misdirect predatory attacks, whereas dark body coloration might provide ultraviolet light protection and dark colour around the eyes might serve an anti-glare function (Ficken et al., 1971; Ortolani, 1999; Kelley et al., 2013; Mueller & Neuhauss, 2014; Kjernsmo et al., 2016).

In many fish, cephalopods and lizards, body colour is mediated through the density and distribution of different types of hormonally regulated chromatophores in the skin (Fujii, 2000; Nilsson Sköld et al., 2013, 2016). Chromatophores are grouped by their colour under white light as xanthophores (yellow), erythrophores (red), leucophores (white), cyanophores (blue) and melanophores (black). Chromatophores can rapidly contract or dilate to allow colour to change within seconds. This permits body coloration to be a highly dynamic signal, which can be modified in response to the momentary intentions of the sender, the proximity of receivers or the perceived threat of predation. In fish, it is not unusual for both males and females to adjust coloration in relation to courtship, aggression and territoriality and in response to the presence and abundance of predators or rivals (Dawkins & Guilford, 1993; Kodric-Brown & Nicoletto, 1993; Bernet et al., 1998; Kodric-Brown, 1998; Svensson et al., 2006; Watson et al., 2014). In some species, aspects of the female reproductive cycle are accompanied by specific coloration. For example, if females spawn repeatedly throughout the reproductive season, nuptial coloration can reflect current receptivity plastically and therefore limit the potential cost of lost crypsis (Rowland et al., 1991; McLennan, 1995). Conversely, in species where females incur a high cost of male harassment or predation risk during copulation, females that have already mated might use prominent coloration to deter male attention (Watkins, 1997; Hager, 2001; Massironi et al., 2005; Tobias et al., 2012).

The regulation of chromatophore-mediated colour change depends on whether it occurs quickly, within seconds or minutes, or slowly, over the season or between life stages. Rapid colour change is a physiological process that involves coordinated aggregation or dispersal of pigment organelles within the chromatophores or that occurs by reorganization of crystals within the chromatophores to shift the reflected wavelength, whereas slower change is a morphological process that alters the number of chromatophores or pigment organelles (reviewed by Nilsson Sköld et al., 2016). Key regulatory hormones in rapid colour change include melanocyte-stimulating hormone (MSH), which elicits dispersal of pigments resulting in a darkening of the skin, and noradrenaline (NA) and melanocyte-concentrating hormone, both of which trigger skin paling. Chromatophores can also be light sensitive, with light inducing either pigment aggregation (Wakamatsu et al., 1980) or dispersion (Iga & Takabatake, 1983). Although chromatophore regulation is well understood at cellular and molecular levels, few studies have examined regulation in relation to the ecology of the whole organism (Nilsson Sköld et al., 2008, 2015).

This begs the question of how dynamic and sexually dimorphic nuptial coloration is regulated. Comparative studies on permanent sexually dimorphic coloration indicate that bright colours might be linked to male-typical androgen levels, which prevent them from being expressed in females (James & Sampath, 2007; Kipouros et al., 2011; Sedláček et al., 2014). Studies on dynamic female nuptial coloration, such as the orange belly shown by two-spotted gobies, Gobiusculus flavescens (Fabricius), have linked it to exposure to melatonin, prolactin and MSH (Nilsson Sköld et al., 2008). Comparative studies on the regulation of dynamic sexually dimorphic nuptial coloration are few but indicate that the dimorphism might be linked to underlying histological differences. For example, the blue eyes characteristic of the nuptial coloration in male three-spined sticklebacks, Gasterosteus aculeatus L., emerge when NA, together with MSH and prolactin, cause the melanophores in the iris to aggregate and reveal the underlying tissue, which is blue in males and pale in females (Franco-Belussi et al., 2018). In the squid Doryteuthis opalescence (Berry), females can display bright and iridescent stripes and patches that are associated with thick layers of iridocytes with numerous lamellae, which are absent in males (DeMartini et al., 2013).

Here, we examine the case of female sand gobies, Pomatoschistus minutus (Pallas), and a conspicuous darkening of the eyes and the surrounding skin, referred to as ‘dark eyes’, which they sometimes display when engaging with a courting male (Kvarnemo et al., 1995; Pedroso et al., 2013). Dark eyes are never observed in males. In females, the intensity of display can vary from a weak line to a full mask and can be switched on or off within seconds.

Although the display is striking, its function remains unclear. Experimental studies suggest that it does not elicit male interest or form part of intrasexual aggression among females, although it is more likely to be shown by rounder females and might therefore be related to imminent spawning (Olsson et al., 2017). Water depth affects ambient light intensity, and the morphology of fish eyes has been shown to possess corresponding adaptations, which in shallow-water species include traits thought to serve an anti-glare function (Lythgoe, 1975; Hunt et al., 2015). During the reproductive season, sand gobies typically inhabit shallow, unstructured, sandy flats, and previous work has indicated that sand goby eyes are light sensitive. The iridophores in the cornea rapidly confer a dark blue-green iridescence, which might reduce intra-optical flare by deflecting sunlight (Shand & Lythgoe, 1987; Lythgoe & Shand, 1989). The darkness of the iris has been shown to be controlled hormonally, paling when exposed to melanocyte-concentrating hormone and NA and darkening when exposed to MSH and adrenocorticotrophic hormone (Nilsson Sköld et al., 2015), but it is not clear how the sexually dimorphic dark eye display is regulated. Furthermore, although melanophores in excised biopsies did not respond to light, live fish transferred from a white to dark background showed darkened skin, especially around the eyes (Nilsson Sköld et al., 2015), implying that light sensitivity is triggered at the organismal level, rather than at the cellular level. Ecologically, dark markings around the eyes have been shown to reduce glare, and dark lines leading from the eyes to the snout to focus line of sight (Ficken et al., 1971; De Broff & Pahk, 2003). A female that carries mature eggs will typically inspect several males before spawning, each of which will show off his nuptial coloration and try to lead her to his nest (Forsgren, 1997). It was therefore hypothesized that dark eyes in spawning-ready females might aid visual acuity, which is likely to be especially important during mate sampling (Olsson et al., 2017).

The purpose of the present study was to examine how dark eyes are regulated to yield a sexually dimorphic display, and the ecological context of the display. In terms of regulation, we asked whether the display is sexually dimorphic owing to (1) a difference in the number of melanophores between males and females, or (2) their response to hormonal exposure. We also examined whether females use dark eyes as an aid to vision, for example when inspecting mates. Specifically, we tested the hypotheses that (3) females are more likely to display dark eyes under bright light, and that (4) females are more likely to display dark eyes in the company of nest-holding males.

MATERIAL AND METHODS

Study organism

The sand goby is a small fish inhabiting coastal waters in northern Europe. During the breeding season, lasting approximately from mid-April to mid-June in our study area, adults migrate from deeper waters into shallow sandy bays, where males build nests under shells or stones and cover them with sand (Hesthagen, 1977). Although both males and females spawn repeatedly during the single breeding season, the operational sex ratio is usually male biased (Kvarnemo, 1994). Male courtship is conspicuous. It consists of fin displays, during which the male shows off his nuptial coloration, in particular an iridescent spot on the dorsal fin, an iridescent band with a black trim on the anal fin, and sooty tail and pectoral fins (Kvarnemo et al., 1995). In contrast, females are coloured cryptically, except for a brief display of dark eyes (Fig. 1 insets), which is more likely to be shown by females carrying mature eggs than by slimmer females (Olsson et al., 2017). When a female has chosen a mate, she deposits her eggs in the nest and leaves the care and guarding of the eggs to the male (reviewed by Forsgren, 1999).

Experimental design

The study was conducted at Kristineberg Marine Research Station (58°15′N, 11°28′E), University of Gothenburg, on the Swedish west coast in May and June 2018. Sand gobies were collected from the adjacent Bökevik Bay using hand trawl, immediately taken to the laboratory, separated by sex and placed in holding aquaria (130 L), which were kept in a greenhouse and exposed to natural daylight. All fish were sexually mature. For males, only individuals in breeding coloration were kept (thus avoiding sneaker morph males; Kvarnemo et al., 2010), and for females, only round females visibly carrying mature eggs were kept. Fish in holding aquaria were fed daily with finely chopped, frozen Alaska pollock, Gadus chalcogrammus Pallas, and brown shrimp, Crangon crangon (L.). All aquaria used during the study were supplied continuously with natural sea water pumped in from a depth of ~7 m.

Hormone experiment

To test whether males and females respond differently to hormone exposure, we subjected in vitro samples to a hormone and colour change assay, according to a protocol outlined by Franco-Belussi et al. (2018). Stock solutions of 1 mM NA (Sigma Aldrich; www.sigmaaldrich.com) and 1 mM α-MSH (ICN Pharmaceuticals, Costa Mesa, CA, USA), were stored at −20 °C and diluted to experimental concentrations of 10 and 5 μM, respectively, in Atlantic cod Ringer solution at pH 7.5 (150 mM NaCl, 5.2 mM KCl, 1.8 mM MgSO₄, 7.0 mM NaHCO₃ and 1.9 mM NaH₂PO₄), immediately before use. Previous work has shown that these hormones at the concentrations used here affect skin melanophores in sand gobies (Nilsson Sköld et al., 2015).

Ten males and ten females, taken at random from the holding aquaria, were measured for total body length to the nearest millimetres (length did not differ between males and females; Student’s unpaired t-test, t_30.8 = 0.44, P = 0.66; mean ± SE: males, 58.7 ± 0.01 mm and females, 59.3 ± 0.03 mm). Female roundness was scored as described by García-Berro et al. (2019), by visual inspection on a scale from one to three in quarter-steps, where one is slim and three very round. Male nuptial coloration was scored in a similar manner, from one to three in quarter-steps, where one is pale and three well coloured (for photographic reference see Supporting Information, Fig. S1). We refer to these scores jointly as the ‘reproductive maturity index’ (RMI). Each fish was placed in a Petri dish partly filled with water and photographed, before being pithed and decapitated. The head was then cut sagittally such that a pair of biopsies, each consisting of half a head with an eye, surrounding skin and half a jaw, was obtained from each fish. One biopsy from each fish was placed in a solution of NA and one in MSH. All samples were incubated in the dark at room temperature for 30 min, after which they were photographed on a light table under standardized light settings using a camera attached to a light microscope (×2.5; Leica M205 C; Leica Microsystems; www.leica-microsystems.com).

We measured the darkness of the skin biopsies from the photographs on a scale from 0 to 255 in Lab colour space, with 0 corresponding to black and 255 to white, using Photoshop as described by Svensson et al. (2005). We measured the darkness of the dorsal half of the iris, the dark line running from the eye to the snout, in addition to the darkness of the areas dorsally and ventrally of the line to serve as reference. We obtained the line darkness and the reference darkness measurements from all 20 fish, and iris darkness measurements from 18 fish (measurements were missing from one male and one female from the NA treatment, because in one case the eye was damaged and in the other case the eye had accidentally become dislodged from the head, making measurements from those biopsies incompatible with the rest of the data). Each biopsy was analysed blindly with respect to fish identity and sex.

For each biopsy, the relative number of melanophores in the iris was estimated by placing a square of fixed size (0.2 mm × 0.2 mm) over a random part of the image of the dorsal part of iris, because this is the more pigmented part (Nilsson Sköld et al., 2015), and counting the number of melanophores contained within the square. If the melanophores were dense, the microscope zoom was used, and the image was magnified to resolve the photograph until individual melanophores could be identified. This procedure was repeated four times, and the mean number of melanophores was calculated. For five males and six females (i.e. one male and one female in the MSH treatment, four males and two females in the NA treatment, and three females in both treatments), the eyes were completely black, making it impossible to count individual cells. Melanophore counts from both treatments were obtained from four females and five males, and there was a significant correlation between the melanophore count in each pair (Pearson’s product–moment, t₇ = 3.88, P = 0.0061, r = 0.83). Given that the number of melanophores is a morphological trait, it is unlikely to be affected by the in vitro exposure to hormones; hence, we pooled data from both treatments in the melanophore count analyses and excluded treatment.

In total, we obtained counts of the number of melanophores in the iris from seven different females (six from MSH treatment and five from NA treatment) and from all ten males (nine from MSH treatment and six from NA treatment). The same procedure for calculating the mean number of melanophores was applied to the dark line running from the eye toward the snout (Fig. 1), in addition to the paler skin directly dorsally and ventrally of the line, the latter two to serve as a reference baseline. Furthermore, a melanophore index was estimated for the line and for the skin dorsally and ventrally of the line (see Supporting Information, Fig. S2). The index is based on a manual assessment of physiological colour change on a scale of one to five, where an index of one describes aggregated pigments and an index of five dispersed pigments. We obtained melanophore counts for the dark line and the dorsal and ventral baseline areas from all biopsies. The melanophore counts were undertaken in a blinded manner with respect to fish identity and sex. All data from the hormonal experiment can be found in Supporting Information, Table S3.

Light experiment

To test whether light intensity affects the likelihood and extent to which females show dark eyes, we set up two experimental aquaria (50 L, filled to 35 L). Each aquarium was furnished with a 3-cm-thick layer of natural sand, sourced from the same bay where the fish were caught, and divided by a perforated clear Plexiglass screen into a smaller and a larger section. The larger section contained a halved clay flowerpot (outer diameter 7 cm) to serve as an artificial nest site. At a height of 44 cm above the water surface and 57 cm above the layer of sand of each aquarium, a light source (ATI Sirius X2; Ati Aquaristik, Hamm, Germany) was suspended, set to either 15 or 100% white light intensity. Opaque partitions ensured that no light from other light sources spilled over to the experimental aquaria. To avoid aquarium bias, we shifted the light strength twice during the study such that one aquarium accommodated 24 replicates at 15% light intensity and 20 at 100% light intensity, whereas the other accommodated 20 replicates at 15% light intensity and 23 at 100% light intensity. We measured absolute spectral irradiance in the experimental aquaria to establish the difference between 100 and 15% light intensity treatments (Fig. 2). The measurements were taken 52 cm below the lamp and 15 cm below the water surface, using a calibrated Jaz spectrometer (Ocean Optics, Dunedin, FL, USA) attached to CC-3-UV-5 cosine-corrected irradiance probe with a 400 μm optic fibre. Both aquaria were supplied with natural sea water (temperature range 10.5–14.0 °C for the duration of the study), with an inlet in the smaller section and an outlet in the larger section.

Within each light treatment, an experimental female was either accompanied by two males or alone. For replicates belonging to the male-presence treatment, two males were introduced to the larger section in the evening, together with two ripe females. These females were confined inside a transparent, perforated container to stimulate male nest-building behaviour. In all trials but ten (four in the 15% intensity and six in the 100% intensity), evidence of nest-building was detected the following morning. The stimulus females were then removed, and a mature (RMI 1.75–3) focal female was introduced into the aquarium. A previous study found that females with RMI ≥ 1.5 may show dark eyes (Olsson et al., 2017). For replicates in the male-absent treatment, the focal female was introduced to an identically furnished aquarium that was empty of other fish. In both cases, the focal female was placed in the smaller section and given ~10 min to acclimate. The fish were then left undisturbed for 30 min while their behaviour was recorded using digital video cameras (Canon Legria HF M56; Ōta, Tokyo, Japan). After completion of the recording, the water temperature was measured, and the male maintaining a position in or close to the nest was identified as the nest-holder. All fish were then captured. Body length was measured to the nearest millimetre, and female RMI was scored in the manner described above (male RMI was not noted in this experiment). At the end of each trial, the layer of sand was smoothed, and the nest was cleared of sand and placed at the centre of the aquarium. After each trial, all fish were released back into the wild, and no fish were re-used.

A total of 88 trials were completed (15% light, males absent, N = 22; 100% light, males absent, N = 23; 15% light, males present, N = 22; and 100% light, males present, N = 21). In one trial (100% light, males present) condensation had formed on the aquarium, rendering the video impossible to analyse; in another (15% light, males present) the female did not appear healthy at the time of the recording; and in one trial (100% light, males absent) the recording failed. These three replicates were removed from analysis, resulting in a total of 85 replicates (15% light, males absent, N = 22; 100% light, males absent, N = 22; 15% light, males present, N = 21; 100% light, males present, N = 20).

There was a slight difference in total length between nest-holding and non-nest-holding males (mean ± SE nest-holders, 62.3 ± 0.02 mm; non-nest-holders, 59.9 ± 0.02 mm; Student’s paired t-test: t₃₈ = 2.07, P = 0.045), but there was no difference in length between males and females (Student’s paired t-test: t_161.8 = −0.007, P = 0.99; mean ± SE females, 61.1 ± 0.02 mm; all males, 61.1 ± 0.01 mm). There was no difference in total length or RMI between females allocated to either light treatment or to male-presence treatment (two-way ANOVA, length: light F_1,81 = 0.32, P = 0.57, male presence F_1,81 = 0.32, P = 0.57, interaction F_1,81 = 1.46, P = 0.23; RMI: light F_1,81 = 0.12, P = 0.73, male presence F_1,81 = 2.11, P = 0.15, interaction F_1,81 = 0.49, P = 0.49). There was no difference in the length of males allocated to either light treatment (one-way ANOVA, light: F_1,78 = 1.73, P = 0.19).

Videos were prescreened for whether dark eyes were displayed or not, assessed by whether a colour change could be detected visually. An independent observer conducted the subsequent analyses of the videos and determined the onset, the duration and the intensity of the display. Intensity was assessed on a scale from one to five, where five is the most intense, calibrated by the recorded displays of dark eyes judged to be the least and most intense. In two trials, two separate instances of dark eye display were observed. In these cases, the onset of the display was taken to be the onset of the first display, the duration as the total duration of both displays, and the intensity as the weighted intensity given the fraction of the total duration each instance represented. Owing to the nature of the treatments, it was not possible to analyse the recordings blind. All data from the light experiment and the spectral irradiance measurements can be found in Supporting Information, Tables S4 and S5, respectively.

Statistics

Hormone experiment

To analyse the contrast between the darkness of the line running from the eye to the snout and the surrounding skin (Fig. 1) and the contrast between the darkness of the dorsal half of the iris and the surrounding skin, we calculated the hormone response as: Δdarkness_line = average (darkness_{dorsal of line}, darkness_{ventral of line}) minus darkness_line, and Δdarkness_iris = average (darkness_{dorsal of line}, darkness_{ventral of line}) minus darkness_iris. Consequently, a higher value of Δdarkness corresponds to a greater contrast (i.e. a darker line or iris against a lighter surrounding skin). We tested the effect of sex and hormone treatment on Δdarkness_line and Δdarkness_iris using mixed-effects linear models, with sex and treatment as fixed effects, fish identity as a random effect and non-significant interactions removed. We also tested the effect of RMI, but because RMI is assessed based on different traits in males and females, we performed separate mixed-effects linear models for each sex, with RMI and treatment as fixed effects, fish identity as a random effect and non-significant interactions removed. In the case of Δdarkness_iris in females, fitting a mixed-effects model resulted in a singularity; therefore, linear regressions for each treatment were used instead.

To assess the response of the melanophores to hormonal exposure, we tested the effect of sex and treatment on the melanophore index. Given that the melanophore index is an ordinal variable, we performed mixed-effects ordinal logistic regression, with sex and treatment as fixed effects, fish identity as a random effect and non-significant interactions removed. We carried out three separate models, for the line and for the skin ventral and dorsal to the line, respectively.

To examine whether the number of melanophores differs between the sexes, for each biopsy we calculated the mean number of melanophores in the line, above and below the line and in the iris. We tested the effect of sex and body length on the number of melanophores using mixed-effects linear models, with fish identity as a random effect. Again, because RMI is based on different traits in males and females, we tested the effect of RMI and total length using mixed-effects linear models separately for males and females and removed non-significant factors.

We examined the effect of hormonal treatment and the effect of sex on the fraction of replicates in which the iris was completely black using Fisher’s exact test.

The output from the mixed-effects models is presented in the Supporting Information (Tables S1 and S2).

Light experiment

We analysed the effect of light intensity and presence of males on the occurrence of female display of dark eyes using the Cochran–Mantel–Haenszel χ ² test, with male presence as stratum, after verifying that the Woolf test for homogeneity of odds ratios across strata was not significant. We performed three separate logistic regressions to examine whether female RMI, female length or the length of the nest-holding male would affect the display of dark eyes.

In trials in which females displayed dark eyes, we examined the onset (i.e. the time elapsed from the start of the trial until dark eyes were first observed) and the duration of the dark eye display in multivariate regressions, separately for males and females. In both cases, the onset and duration of the dark eye display were the dependent variables. For males, the independent variables were the body lengths of the nest-holding and non-nest-holding males, and for females the independent variables were female body length and RMI.

Finally, we examined how male and female traits affected the display of dark eyes. Given that display intensity of dark eyes in females is an ordinal variable, we tested the effects of female traits (body length and RMI) and male traits (body length of nest-holder and body length of non-nest-holder) in separate ordinal logistic regressions. Interactions were not included owing to a high model eigenvalue ratio.

All statistical analyses were performed in R v.4.1.1 (R Core Team, 2021). In the mixed-effects models, P-values were calculated using Satterthwaite’s method implemented in the package lmerTest (Kuznetsova et al., 2017).

RESULTS

Hormone experiment

There was a significant effect of both sex and hormone treatment on Δdarkness_line (mixed-effects linear model, sex: t_18.0 = −3.30, P = 0.004; treatment: t_19.0 = −3.26, P = 0.004; interaction not significant), with MSH producing a darker line than NA and with females showing darker lines than males (Fig. 3). Separate analyses for males and females on the effect of treatment and RMI (i.e. roundness for females and breeding coloration for males) on Δdarkness_line showed that treatment had a significant effect in females but not males, whereas there was no effect of RMI for either sex (mixed-effects linear model, females: treatment t₉ = −3.23, P = 0.01, RMI t₈ = 1.87, P = 0.10; males: treatment t₉ = −1.42, P = 0.19, RMI t₉ = −0.95, P = 0.37; interactions not significant).

Neither sex nor treatment affected Δdarkness_iris(mixed-effects linear model, sex: t_16.7 = 0.98, P = 0.34; treatment: t_17.1 = 0.35, P = 0.73; interaction not significant). There was no effect of either treatment or RMI on Δdarkness_iris, tested in separate analyses for males and females (mixed-effects linear model, males: treatment t_7.6 = 0.93, P = 0.38, RMI t_7.3 = −0.28, P = 0.78; interactions not significant; linear regression, females: RMI (MSH treatment) F_1,8 = 1.09, P = 0.33, RMI (NA treatment) F_1,7 = 1.29, P = 0.29).

There was a significant effect of both treatment and sex on the melanophore index of the line (mixed-effects ordinal logistic regression, treatment β = −2.22, Wald’s z = −3.47, P < 0.001; sex β = −1.05, Wald’s z = −2.02, P = 0.043; interaction not significant), with a higher index for females and MSH treatment (Fig. 4). Given that β is the variable coefficient on a logarithmic odds scale, a unit change in the variable causes the logarithmic odds of the value of the dependent variable to increase by β, or the odds to increase by e^β. Specifically, the odds of the expected index change by e^−2.22 = 0.11 (i.e. decrease by 89% for a male compared with a female) and by e^−1.05 = 0.35 (i.e. decrease by 65% for NA compared with MSH treatment). Treatment, but not sex, also had a significant effect on the melanophore index of the skin dorsal and ventral to the line (mixed-effects ordinal logistic regression, ventral: treatment β = −2.01, Wald’s z = −3.6, P < 0.001; sex β = −0.01, Wald’s z = −0.02, P = 0.98; dorsal: treatment β = −2.64, Wald’s z = −3.4, P < 0.001; sex β = −0.29, Wald’s z = −0.69, P = 0.49; interactions not significant; i.e. the odds of the expected index decrease by 87% for NA compared with MSH treatment ventrally, and by 93% dorsally).

There was no difference in the number of melanophores, either in the line or in the iris, between males and females (mixed-effects linear model, line: sex t₁₈ = 0.68, P = 0.51; iris: sex t_14.9 = −0.40, P = 0.70; Fig. 5). There was also no effect of RMI or total length on the number of melanophores, in either females (mixed-effects linear model, line: RMI t₇ = 0.74, P = 0.49, total length t₇ = 0.024, P = 0.98; iris: RMI t_4.02 = 0.65, P = 0.55, total length t_4.27 = −0.05, P = 0.96; interactions not significant) or males (mixed effect linear model, line: RMI t₇ = 0.70, P = 0.51, total length t₇ = −0.81, P = 0.45; iris: RMI t_7.2 = −0.75, P = 0.48, total length t_6.9 = −0.27, P = 0.79; interactions not significant).

In 14 of 40 biopsies (four females and one male from MSH, five females and four males from NA), the iris was completely black. There was no effect of either treatment or sex on the number of replicates in which the iris was completely black (Fisher’s exact test, P = 0.32).

Light experiment

A total of 21 females showed dark eyes, but there was no effect of light treatment, with or without male presence (Cochran–Mantel–Haenszel test, χ ² = 2.03, d.f. = 1, P = 0.15), on the number of females that displayed dark eyes (Fig. 6).

There was no effect of RMI on the probability of females showing dark eyes, although there was a slight tendency for smaller females to be more likely to display dark eyes (logistic regression, RMI Wald’s z = 1.45, P = 0.15; female length Wald’s z = −1.76, P = 0.08, interaction not significant). In the male-present treatment group, there was no effect of the body length of either the nest-holder or the non-nest-holder male on the probability of females showing dark eyes (logistic regression, nest-holder: length Wald’s z = 0.51 P = 0.61; non-nest-holder: length Wald’s z = 0.45, P = 0.65; interaction not significant).

In trials in the male-present treatment group in which females showed dark eyes, the length of the nest-holder, but not the length of the other male, affected the dark eye display [multivariate regression (dependent variables: display onset and display duration), nest-holder: Pillai_2,7 = 0.67, P = 0.02; non-nest-holder: Pillai_2,7 = 0.18, P = 0.51; interaction not significant]. Linear regressions of nest-holder length run separately for the two dependent variables showed that duration was negatively affected by male length, whereas onset was unaffected (linear regression, duration: nest-holder F_1,10 = 14.1, P = 0.004; onset: nest-holder F_1,10 = 1.1, P = 0.32; Fig. 7). The dark eye display was not affected by either female length or RMI [multivariate regression (dependent variables: display onset and display duration), RMI: Pillai_2,17 = 0.09, P = 0.47; length: Pillai_2,17 = 0.09, P = 0.44; interaction not significant].

Display intensity was also negatively affected by the length of the nest-holder but unaffected by the length of the non-nest-holder (ordinal logistic regression, nest-holder: β = −0.50, Wald’s z = −2.37, P = 0.018; non-nest-holder: β = 0.18, Wald’s z = 1.23, P = 0.22; i.e. the odds of the expected intensity decrease by 39% per millimetre increase in nest-holder length). The display intensity was not significantly affected by female length or RMI (ordinal logistic regression, length: β = −0.10, Wald’s z = −1.56, P = 0.12; RMI: β = −0.85, Wald’s z = −0.47, P = 0.64).

DISCUSSION

During courtship, female sand gobies are known temporarily to display a dark mask across the eyes and running down towards the snout, which is distinct from any aspect of male coloration. In the present study, we showed that this in vivo sexual colour dimorphism was consistent with in vitro differences between the sexes in the melanophore response to hormonal exposure. Specifically, we showed that in vitro exposure to MSH darkened the line running from the eye to the snout and that this treatment affected females but not males, whereas maturity, as estimated by the RMI, had no effect on darkness. We also showed that MSH in females elicited a stronger dilatation of the melanophores in the line compared with males. Conversely, we found no effect of sex, hormone treatment or maturity on the darkness of the iris, and we found no difference between the sexes in the number of melanophores in the iris. In vivo, we found no effect of light intensity or the presence of potential mates on female propensity to display dark eyes; however, when females showed dark eyes, smaller males elicited a stronger response in terms of the duration and intensity of the display.

The regulatory mechanisms behind sexually dimorphic colour signals have thus far received limited attention, despite the overall importance placed on nuptial coloration for mate choice. In comparison to two studies on fish and squid, which identified histological differences between males and females (DeMartini et al., 2013; Franco-Belussi et al., 2018), we showed that in sand gobies the female-specific dark eye display is not associated with a sex differences in the number of melanophores, but in the physiological response of the melanophores to hormonal exposure. To our knowledge, this is the first study to show how a dynamic, sexually dimorphic colour change results from a sex-specific physiological response to hormonal regulation, rather than from underlying morphological differences.

Our method of using the entire head of the fish for our analyses allowed us to examine the entire area of interest, comparing the effect of treatment separately for the iris and the line. However, by not extirpating the eyes fully from the head, it is possible that remaining tissue prevented the medium from reaching the eye and effecting a physiological colour change. This might explain why we found no effect of hormones on the darkness of the iris, compared with previous work on sand gobies showing that both melanophores and erythrophores of the eyes contract when exposed to either NA or melanocyte-concentrating hormone, but dilate in the presence of MSH and adrenocorticotrophic hormone (Nilsson Sköld et al., 2015).

Generally, eyes in fish are cryptically coloured or masked by disruptive colours (Nilsson Sköld et al., 2016), but rapid and conspicuous colour changes have been observed in many species, although what is being signalled varies widely. For example, despite possessing a bright red anterior belly, throat and lower head, courting male three-spined sticklebacks use conspicuous blue eyes to attract females (McLennan, 1995; Flamarique et al., 2013). In both Nile tilapias, Oreochromis niloticus (L.), and Atlantic salmon, Salmo salar L., the eyes of dominant individuals are pale, compared with subordinates that display dark eyes, especially after aggressive encounters (Suter & Huntingford, 2002; Volpato et al., 2003; Vera Cruz & Brown, 2007). The eyes of tilapias also darken when exposed to stress (Freitas et al., 2014). Likewise, female bluegill sunfish, Lepomis macrochirus Rafinesque, approaching a nest-holding male display dark eyes and barring on the body, possibly to lower male aggression, as has been shown in the related pumpkinseed sunfish, Lepomis gibbosus L. (Colgan & Gross, 1977; Dominey, 1980, 1981). Conversely, in guppies, Poecilia reticulata Peters, black eyes signal aggression and dominance (Martin & Hengstebeck, 1981; Magurran & Seghers, 1991) and can provoke attacks from competitive conspecifics (Heathcote et al., 2018). The eye-bar, a prominent dark facial stripe, in Astatotilapia burtoni (Günther) is also related to male dominance and aggression (Muske & Fernald, 1987). Finally, it is unclear what function is served by the red fluorescent eyes displayed by the black-faced blenny, Tripterygion delaisi Cadenat & Blache, but it might aid in foraging (Wucherer & Michiels, 2014).

Why do female sand gobies display dark eyes? Although sexually dimorphic traits, including eye colour, are often assumed to be ornamental, they can also result from different ecological selective pressure on males and females or be involved in female-specific signalling (Shine, 1989; Schlupp, 2018). If dark eyes are ornamental, it would be reasonable to expect males to prefer dark-eyed females and for female displays to be directed toward larger males, because females are known to prefer larger males (Forsgren, 1992; Kvarnemo & Forsgren, 2000). However, previous work found no evidence that males respond to displays of dark eyes (Olsson et al., 2017), and we found no effect of male presence or male size on the probability of the dark eye display. Intriguingly, however, the displays documented here were more intense and longer lasting when directed toward smaller males, which is contrary to expectations both if dark eyes are ornamental and if dark eyes would serve to lower male aggression. If dark eyes indicate aggression, especially to ward off other females in competition for a preferable mate or nest site, we might expect a random pattern of pre-emptive displays, because our experiments included only one focal female. Yet, previous work failed to find a link between dark eye displays and female aggression, either when provoked using mirrors or when two females and one male interacted freely (Olsson et al., 2017). If dark eyes are indicative of a readiness to spawn, as suggested in earlier work (Olsson et al., 2017), we would expect maturity to affect the propensity for display. In the present study, we found no such effect, but we deliberately picked relatively round females in order to examine other factors and might not have had enough variation in roundness to identify an effect.

We hypothesized that dark eyes might aid vision and possibly serve an anti-glare function (Ficken et al., 1971; Shand et al., 1987; Lythgoe et al., 1989; De Broff et al., 2003), which would be especially important when sampling mates. We thus expected females to be more likely to show dark eyes in high-intensity light and in the presence of nest-guarding males. However, we found no evidence that either bright light or the presence of a male with a nest affected the incidence of dark eyes. Natural light in shallow sandy bays (such as the bay where we collected the fish used in the present study) can reach higher levels than we had in our 100% light intensity treatment (Häder, 2001). Nevertheless, the experimental light levels were sufficient to induce the dark eye display in females. Furthermore, we had a marked difference in light intensity between our treatments, and if dark eyes were displayed in response to high-intensity light, we should have seen an effect in that direction. Instead, the relative incidence of dark eyes was slightly, albeit not significantly, higher in the low light treatment, as can be seen in Figure 6.

Although dark eyes are conspicuous and therefore appear to be a deliberate cue with a specific function, it is nevertheless possible that the dark eyes are a by-product of some other process. For example, the rapid darkening of the meninx sometimes observed in guppies has been hypothesized to serve as ultraviolet light protection, but on testing the hypothesis it was instead found to correspond to stress levels (Gibson et al., 2009).

In summary, we show that the display of dark eyes in female sand gobies is hormonally regulated and distinct from males owing to a sex-specific physiological response of the melanophores, which become more dilated in females, consistent with the dark colour around the eyes observed in females but not in males. We also show that the sex-specific response does not arise from underlying differences in the number of melanophores in males and females, but in the physiological response of the melanophores to hormonal exposure. We find no evidence that dark eyes in female sand gobies are ornamental or serve an anti-glare function in bright light.

SUPPORTING INFORMATION

Additional Supporting Information may be found in the online version of this article at the publisher’s web-site:

Figure S1. Reproductive maturity index (RMI) of sand goby (Pomatoschistus minutus) males, ranging from 0.75 to 3.00. The photographed males were caught at the peak of the breeding season, near Kristineberg Marine Research Station, University of Gothenburg, Sweden, in early to mid-June 2014 or 2018, or near to Tvärminne Zoological Station, University of Helsinki, Finland, in late June 2014. Fish pictured from above were photographed when alive, whereas fish pictured from the side were photographed after they had been killed. The black pigmentation becomes a little more evident in the latter group. For this study, all scoring of RMI was done on live males, placed in a transparent plastic cup under good light, to be able to see the black coloration on the ventral and anal fins, in addition to the iridescent blue band on the anal fin, which is not readily seen in these photographs.

Figure S2. Physiological colour change in the sand goby (Pomatoschistus minutus). Exposure to melanocyte-stimulating hormone (MSH) causes the chromatophores to dilate to give a darker appearance, whereas exposure to noradrenaline (NA) causes the chromatophores to contract and give a lighter appearance. Images on the left show the studied part of the body surface and images on the right show the skin in magnification (millimetre scale included in the corner of each image). Photographs of fish caught during the course of the present study were taken by H. Nilsson Sköld.

Table S1. Output of mixed-effects model pertaining to: (1) darkness (overall effect of treatment and sex); darkness for (2) females and (3) males separately [effect of reproductive maturity index (RMI) with treatment as covariate]; (4) darkness (overall effect of treatment and sex); (5) darkness for males separately (effect of RMI with treatment as covariate; effect of RMI on females was fitted with linear regression separately for each treatment owing to singularity when fitting models and is reported in main text); (6) number of melanophores in the line (overall effect of sex and total length); number of melanophores separately for (7) females and (8) males (effect of RMI with total length as covariate); (9) number of melanophores in the iris (overall effect of sex and total length); number of melanophores in the iris separately for (10) females and (11) males (effect of RMI with total length as covariate) and (12) the number of melanophores (overall effect of sex and position: in the line, ventral and dorsal to the line). Models with interactions are shown on the left and models without interactions on the right. Categorical variables [sex: females or males; treatment: melanocyte-stimulating hormone (MSH) or noradrenaline (NA); position: dorsal, line or ventral] are ordered alphabetically such that the baseline case is that of female, MSH and dorsal. Random factor and residual variance are given within dashed lines.

Table S2. Output of ordinal mixed-effects model pertaining to the melanophore index (MI) of: (1) the line; (2) the skin dorsal to the line; and (3) the skin ventral to the line (overall effect of sex and treatment). Models with interactions are shown on the left and models without interactions on the right. Categorical variables [sex: females or males; treatment: melanocyte-stimulating hormone (MSH) or noradrenaline (NA)] are ordered alphabetically such that the baseline case is that of female and MSH. Threshold estimates, and random factor and residual variance given within dashed lines.

Table S3. Data from hormone experiment. Table headings refer to fish identity (ID), hormone treatment [melanocyte-stimulating hormone (MSH) or noradrenaline (NA)], sex [male (M) or female (F)], total length (TL; in millimetres), reproductive maturity index (RMI), and darkness and number of melanophores (#mel) in samples 1–4 for the iris and the dark line. The black eyes column refers to replicates in which no melanophore counts were possible.

Table S4. Data from light intensity treatment. Table headings refer to the replicate number, water temperature (Temp), treatment (light intensity and male presence), total lengths [nest-holding male (NH), non-nest-holding male (non-NH) and female], female reproductive maturity index (RMI), display time and display intensity, including two replicates in which the female showed dark eyes twice.

Table S5. Irradiance measurements (in watts). Irradiance was measured 52 cm below the lamp and 15 cm below the water surface. The grey lines show irradiance in 100% light intensity and the black lines in 15% light intensity, while the continuous lines refer to the male compartment and the dashed lines the female compartment.

ACKNOWLEDGEMENTS

We thank Bengt Lundve and other staff at Kristineberg for support and encouragement during the study, Leon Green for kindly lending us his ATI Sirius X2 aquarium lamps, and Laima Bagdonaitė for video analysis. We also thank two anonymous reviewers for helpful comments and advice. The study complied with all national and international ethical guidelines and was carried out under permits Dnr 205-2013 from the Swedish Board of Agriculture and iDnr 001471 from the Ethical Committee for Animal Research in the Gothenburg Board of Animal Research. The authors have no conflicts of interest to declare. This study was funded by Wåhlströms memory foundation (HNS) and the Swedish Research Council, grant numbers 2016-03343 and 2020-04992 (CK).

DATA AVAILABILITY

All data are available in the Supporting Information.

REFERENCES