More than meets the eye: Predator-induced pupil size plasticity in a teleost fish

1. Most animals are visually oriented, and their eyes provide their ‘window to the world’. Eye size correlates positively with visual performance, because larger eyes can house larger pupils that increase photon catch and contrast discrimination, particularly under dim light, which have positive effects on behaviours that en hance fitness, including predator avoidance and foraging. 2. Recent studies have linked predation risk to selection for larger eyes and pupils, and such changes should be of importance for the majority of teleost fishes as they have a pupil that is fixed in size (eyes lack a pupillary sphincter muscle) and, hence, do not respond to changes in light conditions. 3. Here, we quantify eye and pupil size of individual crucian carp, a common fresh water fish, following controlled manipulations of perceived

Such predator-driven trait changes can become canalized and genetically fixed within populations and species, and thus always expressed, regardless of prevailing risk of predation. However, in nature, environmental conditions are inherently dynamic, and the risk of predation is often spatially and temporally variable, which may favour the evolution of phenotypic plasticity in traits that decrease predation risk (Pigliucci, 2001;Tollrian & Harvell, 1999).
Via plasticity, prey can thus fine-tune anti-predator traits to increase survival chances when predators are present, but also save important expression and maintenance costs in their absence (Pigliucci, 2001).
Visually guided behaviours, such as finding food and detecting potential mates and/or predators, are principal drivers of the evolution of the eye, and have produced a plethora of eye morphologies and visual strategies across taxa and environmental gradients (Banks et al., 2015;Brischoux, Pizzatto, & Shine, 2010;Hall & Ross, 2007;Land & Fernald, 1992;Land & Nilsson, 2012;Schwassmann & Kruger, 1966). The size of the pupil is of fundamental importance for visual performance as the pupil area determines the light-gathering capacity, and, hence, contrast sensitivity and acuity (Caves, Sutton, & Johnsen, 2017;Land & Nilsson, 2012;Veilleux & Kirk, 2014). The size of the pupil is especially important under dim light conditions, and the vast majority of vertebrate taxa are capable of enhancing visual performance via dynamic control of the pupil size in response to changes in light levels using the muscle of the iris (Douglas, 2018;McDougal & Gamlin, 2015). Recent studies have linked selection for large eye/pupil size to e.g. dim light habitats and nocturnal behaviour (Martinez-Ortega, Santos, & Gil, 2014;Schmitz & Wainwright, 2011;Warrant, 2004) or to competition  and predation risk per se (Freund & Olmstead, 2000;Glazier & Deptola, 2011;Møller & Erritzoe, 2010;Nilsson et al., 2012). Such changes may be of particular importance in the many teleost fish that lack a pupillary sphincter muscle and, hence, have a pupil that is fixed in size and not responsive to ambient light intensity (Douglas, 2018;Helfman et al., 2009). Yet, while numerous studies have unravelled how predation risk can affect prey eye and pupil size, both at micro-and macro-evolutionary scales (Banks et al., 2015;Beston, Dudycha, Post, & Walsh, 2019;Land & Nilsson, 2012;Møller & Erritzoe, 2010;Nilsson et al., 2012), few have examined whether individuals can implement changes to eye morphology traits in response to environmental variation within their lifetime (but see Ab Ghani, Herczeg, & Merilä, 2016;Svanbäck & Johansson, 2019 for examples of predator-induced plasticity in overall eye size). The high energetic costs of neural tissue make the eye one of the most energetically expensive organs in the vertebrate body (Laughlin, de Ruyter van Steveninck, & Anderson, 1998;Moran, Softley, & Warrant, 2015;Niven & Laughlin, 2008). By adopting an inducible strategy in the visual system, prey should enjoy a closer phenotype-environment match where investments into eye and pupil size are tuned to the prevailing level of predation risk and where unnecessary costs of expression thus can be saved in the absence of predators (Tollrian & Harvell, 1999).
Here, we tested for predator-induced plasticity in eye and pupil size in a teleost fish, the crucian carp Carassius carassius, a common freshwater fish that occupies small and shallow lakes and ponds in temperate regions across the northern hemisphere. Crucian carp constitute a well-established model system to study phenotypic plasticity and, particularly, inducible morphological defences (Brönmark & Miner, 1992;Brönmark & Pettersson, 1994;Hulthén et al., 2014). Previous field and laboratory studies have shown that crucian carp respond to chemical cues released by predators, such as pike Esox lucius, by an increase in body depth (Brönmark & Miner, 1992;Brönmark & Pettersson, 1994). The morphologically defended phenotype constitutes less desirable prey for gape-limited predators (Nilsson, Brönmark, & Pettersson, 1995), and the deep body improves escape performance via enhanced locomotor capacity (Domenici et al., 2008). However, many prey species also respond behaviourally to increased predation risk by, for example, changes in diel activity patterns (Bakker, Reiffers, Olff, & Gleichman, 2005;Lima & Dill, 1990;Monterroso, Alves, & Ferreras, 2013;Reebs, 2002;Weiss, 2018). When prey are confronted by visually oriented and diurnal enemies, a shift towards nocturnal activity would directly reduce encounter rates. However, for previously diurnal species, predator-induced increases in nocturnality can cause mismatches between eye morphology and the new and novel visual environment and thus interfere with, for example, foraging efficiency. An increase in nocturnality may thus select for facultative and rapid adaptations of the visual system, enabling night-time activity. Such changes in diel rhythms have recently been suggested as the selective agent driving evolutionary shifts in eye size among populations of the Trinidadian killifish Rivulus hartii . Furthermore, a recent study demonstrated that selection pressures on activity patterns have been the main drivers for evolutionary divergence in the mammalian eye (Baker & Venditti, 2019).
In this study, we quantified predator-induced changes in eye and pupil size at the level of individuals among mature, wild-caught and previously predator-naive crucian carp. We also examined diel activity patterns in crucian carp held in the presence or absence of predatory pike. We predicted that individuals exposed to perceived predation risk would develop larger eyes capable of holding a larger pupil. The increased light-gathering capabilities should result in enhanced contrast detection and visual range, and thus aid in predator detection. We also predicted that crucian carp should shift activity patterns towards a higher degree of nocturnality in the presence of pike, a strategy assumed to be adaptive as it would directly reduce the risk of capture from diurnal and visually oriented pike predators (Skov & Nilsson, 2018). Finally, to shed light on how plastic changes to pupil size may affect visual performance, we modelled how the observed trait changes affect maximum detection distances of objects of varying size under diverse light regimes.

| Fish and maintenance
We used wild-caught and previously predator-naive crucian carp were externally covered with a black/blue plastic film. Each experimental tank held four (Experiment 1) or five (Experiment 2) crucian carp individuals and the experiments were initiated when fish were introduced to one of the aquarium compartments. The aquaria water was continuously aerated and filtered (one filter per replicate tank). Crucian carp were fed a mixture of frozen Daphnia, chironomids and shrimp five times weekly corresponding to a ratio of approximately 5% of the total body mass in each replicate tank.
A single pike (Experiment 1: n = 12, standard length: 31-41 cm, Experiment 2: n = 7, standard length 23-31 cm) were introduced to the other compartment in aquaria used for the predator treatment (pike treatment aquaria assigned by random permutation).
Predatory pike received a strict crucian carp diet and were fed on a weekly basis during the acclimatization and experimental period.

| Experiment 1 -Eye plasticity
Trait changes in overall eye and pupil size at the level of individuals were quantified from digital photographs. We photographed all crucian carp prior to treatment exposure and once again when the

| Image analyses
From the digital images we extracted morphological variables using the image analysis software ImageJ v. 1.49 (https://imagej. nih.gov/ij/). We measured the widest part of the eye and pupil at the horizontal plane (Beston, Wostl, & Walsh, 2017;Svanbäck & Johansson, 2019). In addition, we measured standard length (the distance between the tip of the snout to the end of the last scale anterior to the caudal fin) of all fish to account for potential body size differences and enable calculation of relative eye and pupil size.

| Experiment 2 -Diel activity
We examined the effect of predation risk on diel activity patterns in crucian carp in a behavioural experiment. After 166-176 days of treatment exposure, the diel activity patterns of individual fish were recorded within each aquarium. During the diel activity trials we employed different light regimes simulating natural conditions: (a) day (7.00-19.00; ~540 lx), (b) evening (19.00-20.58; ~18 lx), (c) night (21.00-5.00; 0 lx) and (d) dawn (5.00-7.00; ~18 lx). To allow monitoring of fish activity also under low light conditions, we used digital infrared scouting cameras (UOVISION UV572 12 MP HD, infrared wavelength: 960 mm) horizontally facing the centre of the crucian carp compartment. Some fish species have evolved a visual sensitivity for near-infrared light (Enright et al., 2015). However, pilot experiments in our model system showed that crucian carp did not respond behaviourally to the IR light during dark hours, i.e. we could not detect any behavioural response upon the onset of the IR light. The camera was set to record for 1 min each hour for 24 hr, starting at 9.00. All fish were fed 24 hr before each trial started to standardize hunger levels prior to activity measurements. Pike in the predator treatment were fed one crucian carp the day before we recorded activity. A grid (4 × 4 squares) was taped on the front of each aquarium to enable quantification of fish activity. From the recordings we quantified the total number of squares entered by each individual during 1 min at 12 different time points during 24 hr (day: 08.00, 11.00, 14.00, 18.00; twilight: 05.00, 06.00, 19.00, 20.00; night: 01.00, 03.00, 21.00, 23.00), resulting in four time points per light condition). In case of mortality in experimental tanks, fish were immediately removed and replaced with a similar-sized individual so that the initial density level was held constant throughout the experiment.

| Modelling of visual range
The maximum distance for visual detection of dark objects was calculated according to theory developed for aquatic vision (Nilsson, Warrant, & Johnsen, 2014). The pupils were assumed to be circular, and we used the estimated means of the pupil diameters from each treatment group (predator-exposed or control), although we modelled visual performance for the complete pupil range of 0-4 mm. Because we used a camera calibrated to provide radiance per nm of green light across the spectral band 500-600 nm, the resulting radiances had to be multiplied with the spectral absorption of visual opsins to assess the photon flux available for vision. We used a Govardovskii template (Govardovskii et al., 2000) for an opsin peaking at 550 nm, and assumed additional spectral filtering by 2 m of lake water [clear site at Lake Victoria (Seehausen, vanAlphen, & Witte, 1997)]. The resulting photon flux (photons s −1 m −2 sr −1 ) values for bright daylight available for vision at 2 m depth were 1.93 × 10 21 for upward viewing, 3.06 × 110 20 for horizontal viewing and 9.68 × 110 19 for downward viewing. For twilight and starlight conditions, we assumed an intensity reduction of 4 and 8 log units respectively (Nilsson, Warrant, & Johnsen, 2014). To assess the effect of pupil changes on visual performance we calculated the response, R, on the visual range, r, of a change in pupil area, A, as R = (δr/r)/(δA/A).

| Data treatment and statistical analyses
Statistical analyses were performed using SPSS v. 23.0 for Mac OS X (SPSS Inc.).

| Eye plasticity
We used a GLMM to analyse the effect of predator presence/ absence on eye morphology. The change in horizontal (0°) size Moreover, we tested for differences in relative eye and pupil size, using the post-treatment measurements as dependent variables, without and with standard length as a covariate respectively. Tank identity was used as a random factor, nested within treatment to compensate the degrees of freedom according to the experimental design. All data met the assumptions of normality and equality of variances.

| Diel activity
We used a general linear mixed repeated measures model (GLMRM) to test if pike exposure induces a shift towards nocturnality in crucian carp. The mean activity for each time point (n = 12 time points) and aquaria, calculated as the total sum of entered squares within a single aquarium divided by the total number of fish per aquaria (n = 5), was used as response variable. Treatment (predator-exposed or control) was used as a between-subject factor to examine the factor of primary interest (time × treatment). Our data did not meet the assumption of sphericity, and were therefore corrected by the conservative Greenhouse-Geisser procedure (Greenhouse & Geisser, 1959).
Analysing the individual change in pupil size over the course of the experiment revealed that predator exposure induced larger pupils in crucian carps (F 1,23.93 = 15.03, p < 0.001, Figure 1).

| Diel activity
During the 24-hr period of activity recordings, predator-exposed fish were less active than control fish, and activity patterns differed significantly between treatments over the diel cycle (time × treatment interaction term: F 3.204,38.450 = 2.909, p = 0.044), with predator-exposed fish being more active during darkness whereas control fish demonstrated activity peaks during daylight conditions and were relatively less active during low light conditions ( Figure 2).

| Modelled visual range
Modelling the effect of the observed increase in pupil diameter on visual performance revealed a corresponding increase in the visual detection range. The pupil diameter increases by 2.52% (area increases by 5.1%) and depending on light conditions and target size this extends the visual range by 0.28%-2.0% (see Figure 3; Table S1).
The gain in visual range depends only marginally on viewing direction but has a strong dependence on light intensity and on the size of the objects that are visually detected. The strongest performance gain (response) is found for detection of very small targets (prey) in dim light. The results are robust for major variation of input assumptions to the modelling (Table S2). Even though the visual range depends to some degree on water quality (beam and background attenuation coefficients), this has very little effect on the calculated response values.

| D ISCUSS I ON
Our results show that perceived predation risk can induce plastic changes in the vertebrate eye, and also suggest behavioural changes in response to ecological shifts in predator communities. Crucian carp exposed to a live predator (elevated perceived risk) increased pupil but not overall eye size, and also showed a higher degree of nocturnal activity as compared to fish held in the absence of predators.
Next, we applied a theoretical model to shed light on the adaptive value of such trait changes and showed that inducing a larger pupil is a specialization that enhances visual performance, specifically under dim light conditions. Our results thus imply that predator-induced plasticity in a key eye trait may act in concert with behavioural shifts to enable crucian carp to increase survival chances and maintain As a principal sensory organ, the eye provides immediate information on the surrounding environment. The selective advantages of using light as a source of information relate to habitat-specific properties, and, hence, numerous solutions and adaptions to obtain adequate visual information are found in nature (Land & Nilsson, 2012). Eye size is positively correlated with visual acuity and limits visual sensitivity by setting the maximum possible size of the pupil; pupil size ultimately constrains the light-gathering capacity of the eye (Land & Nilsson, 2012). The visual conditions of freshwater habitats are characterized by high absorption and scattering of downwelling light (Holopainen, Tonn, & Paszkowski, 1997), where a larger pupil size significantly improves contrast detection and visual range (Caves, Sutton, & Johnsen, 2017;Nilsson et al., 2012;Veilleux & Kirk, 2014). However, with a few exceptions, teleost fish lack a sphincter pupillae muscle (Douglas, 2018;Helfman et al., 2009) and therefore cannot autonomically regulate the amount of light reaching the retina by changing pupil size in accordance to changes in ambient light. The capability of increasing pupil size in response to changes in perceived predation pressure should hence be adaptive. Relatively large pupils is of particular importance for animals living in environments with poor light conditions, where contrast detection is of more significance than resolution (Land & Nilsson, 2012;Warrant, 2004). Perhaps the most striking example is provided by the giant squid Architeuthis sp. which has evolved spectacularly large eyes and pupils, up to three times larger than any other animal (Land & Nilsson, 2012), as an adaptation for early detection and successful avoidance of predatory sperm whales in the dim light of the deep-sea environment . Moreover, light deterioration in turbid and brown water results in shorter detection and escape distances in common roach Rutilus rutilus attacked by pike, and, hence, pike demonstrate higher attack success and prey fish a lower survival probability in poor light conditions (Ranåker et al., 2012). However, both prey and predators suffer much shorter visual detection distances in dim light.
To evaluate how changes in pupil size affect visual performance in crucian carp we used a model developed for calculating visual performance in water from data on pupil diameter and ambient light intensity (Nilsson, Warrant, & Johnsen, 2014). One important component of the model is the law of diminishing returns, i.e. the marginal value of a relatively larger pupil size decreases with increased pupil size. This general phenomenon is an important limiting factor for eye size in aquatic animals (Nilsson, Warrant, & Johnsen, 2014), but also a reason why small fish larvae have disproportionally large eyes. The observed change in F I G U R E 3 Model effects of pupil size change on visual range (left column) and performance response (right column). Visual range is evaluated as the maximum distance at which a circular black target can be detected with different pupil diameters, and performance response as the ratio between pupil area (as it is linearly related to the observed single-axis change in pupil dimension) change and the resulting change in visual range. A performance response value of 1 implies that a 1% increase in pupil diameter results in a 1% change of the visual range. From top to bottom, the panels display the effects of viewing direction, light condition and target size. The case plotted with a solid line is the same in all panels: horizontal viewing at twilight of a 1 cm target. Dashed lines show the effect of increasing or decreasing indicated variables. The width of the red vertical bar indicates the span of observed pupil dimensions from crucian carp in presence and absence of a predator. All values are calculated for 2 m depth in a lake, although values remain very similar within the possible depth range of 0-3 m (see Table S2) crucian carp pupil area is 5.1% and corresponds to a symmetric change in pupil diameter by 2.5%. This improves the visual range with 0.28%-2.0%, depending on light conditions and visual target size ( Figure 3; Table S1). These seemingly small changes should be viewed in light of the cubic relationship with volume visibility. A 2.5% increase in visual range is associated with a 6.1% increase of the water volume that can be scanned to detect predators and prey, a significant increase that should be of importance for survival and growth opportunities. For example, Brandon, James, and Dudycha (2015) showed that eye size is under selection in a natural population of Daphnia, and also that changes in eye size have clear fitness consequences; small increases (1%) in the eye diameter of Daphnia obtusa increase reproductive outputs by 20%.
Our calculations of the functional response, i.e. how much performance is gained by an incremental increase in pupil area, also provides important insight into the most likely reasons for increasing the pupil area (right panels of Figure 3). The response is almost the same for all viewing directions but increases strongly with lower light intensities and smaller target sizes. This suggests that the predator-induced change towards a nocturnal activity pattern and the resulting need for efficient foraging in dim light can be related to the plastic increase in pupil area. The improvement in detecting larger objects, such as potential predators, in bright light is only marginally increased, indicating that the larger pupil may be a secondary effect of the more crepuscular/nocturnal lifestyle adopted by crucian carp in the presence of predators.
Previous studies have linked the evolution of larger overall eye size to predation risk. For example, Ab Ghani et al. (2016) found that that three-spined sticklebacks Gasterosteus aculeatus show plasticity in eye size and induce relatively larger eyes when reared in presence of chemical cues from predatory Eurasian perch Perca fluviatilis. Furthermore, juvenile perch have recently also been shown to express plasticity in overall eye size as a response to predation risk (Svanbäck & Johansson, 2019). However, in our study, we found no evidence of predator-induced plasticity in overall eye size. This lack of response in overall eye size among predatorexposed fish can be due to various reasons. Phenotypic plasticity and capability of trait modulation can vary substantially over ontogeny (Fawcett & Frankenhuis, 2015;Hochberg et al., 2011;Hoverman & Relyea, 2007;Meuthen et al., 2018), and while the sticklebacks and perch in the studies mentioned above were exposed to chemical cues from predators from the onset of the larval/juvenile stage (Ab Ghani et al., 2016;Svanbäck & Johansson, 2019), we performed our experiments on larger and sexually mature fish (the majority of experimental subjects showed fully developed gonads). Moreover, as pigmented eyes with high contrast against the background can increase the risk of detection by predators, small eyes may have been suggested to be favoured under predator-driven selection (Beston et al., 2017;Svanbäck & Johansson, 2019). Furthermore, as eyes are energetically costly to develop and maintain (Moran, Softley, & Warrant, 2015), the costs associated with changes to pupil size may be significantly lower than investment into construction and maintenance of larger eyes, especially, as an increased eye size ultimately requires a reconstruction and enlargement of the orbit part of the cranium (Goatley, Bellwood, & Bellwood, 2010;Rohner et al., 2013).
A predator-induced change only in pupil size while maintaining overall eye size could be an efficient and fast-acting strategy to enhance vision in dim light. However, for teleost fish this strategy is limited as pupil diameter typically is only marginally smaller than lens diameter, and light entering the eye outside the lens will degrade the image (Douglas, 2018;Land & Nilsson, 2012). Prey adopting inducible defence strategies are often exceptionally vulnerable to predation prior to the expression of the defended phenotype (Holopainen, Tonn, et al., 1997), where rapid increases of pupil size followed by altered diel activity patterns may serve to increase survival before additional anti-predator traits have been expressed.
The behavioural experiment showed that the presence of a natural predator alters the diel activity patterns in crucian carp.
While non-exposed fish maintained a typical diurnal diel rhythm with a peak of activity around noon, pike-exposed crucian carp both reduced general activity and distinctly shifted to a nocturnal activity. This is in line with earlier field studies, where introduction of predatory perch and pike to a natural pond resulted in crucian carp residing in inshore habitats and shifting to nocturnal behaviour (Tonn, Paszkowski, & Holopainen, 1989). Such changes in activity patterns and habitat use are shared across different taxa and a common response in prey, particularly if reliable cues regarding the prevailing risk are present (Lima & Dill, 1990). For crucian carp, a nocturnal lifestyle, enabled by phenotypically plastic pupil enlargement, would lead to reduced encounter rates with typically diurnal pike predators (Skov & Nilsson, 2018). The change to crepuscular/nocturnal activity is expected to have a dramatic effect on predator vision (see Figure 3). For pupil diameters of 5-7 mm (the pike in our experiments), the visual range shrinks to about half from sunlight to twilight, and is further reduced to 12% at starlight.
Compared to sunlight this implies that the water volume covered by vision is reduced with a magnitude of eight times in twilight and with a magnitude of 500 times in starlight. Consequently, the risk of swimming into the detection range of an ambush predator is massively reduced by changing to a more crepuscular or nocturnal activity.
Phenotypic plasticity must theoretically involve costs; if there are no associated costs, trait fixation would be the evolutionary stable strategy (Auld, Agrawal, & Relyea, 2010;Dewitt, Sih, & Wilson, 1998;Tollrian & Harvell, 1999). Empirically, however, costs of plasticity have been shown to be elusive and weak (Van Buskirk & Steiner, 2009), and even though vision is a metabolically expensive sensory system to develop and maintain (Laughlin, de Ruyter van Steveninck, & Anderson, 1998;Moran, Softley, & Warrant, 2015), specific energetic costs associated with investment into a larger pupil are unclear. However, a pupil that is (too) large would increase the risk of direct sunlight entering through a gap between the pupil margin and the lens. Such straylight would seriously compromise contrast sensitivity in exposed parts of the retina. The predator-induced pupil size in crucian carp might therefore not be advantageous when predators are absent and day-time foraging is preferred. In addition, the typical diurnal activity patterns of crucian carp in predator-free environments (Tonn, Paszkowski, & Holopainen, 1989), together with our results, indicate that crucian carp are trading-off their optimal diel activity pattern against safety from predation. Earlier experiments have confirmed that predation risk can reduce the growth rate of crucian carp in the field (Tonn, Paszkowski, & Holopainen, 1992), and numerous studies have shown the importance of light and visual performance to successful detection and attack of prey in planktivorous fish species (Confer et al., 1978;De Robertis, Jaffe, & Ohman, 2000;Hairston, Li, & Easter, 1982;Varpe & Fiksen, 2010).
In our experiment, we found no differences in overall body size between predator-exposed and unexposed individuals, i.e. no evidence of foraging costs associated with the shift in diel activity patterns under these experimental conditions.
In summary, our results demonstrate predator-induced changes in pupil size and diel activity patterns in crucian carp. Coupled with plastic changes in body morphology (Brönmark & Miner, 1992), physiology (Holopainen, Aho, et al., 1997) and behaviour (Höglund et al., 2005;Hulthén et al., 2014;Pettersson, Nilsson, & Brönmark, 2000), our findings highlight that crucian carp have evolved phenotypic plasticity in a broad suite of traits that, when combined, produce an adaptive, integrated anti-predator phenotype. Being a teleost lacking a sphincter pupillae muscle of the iris, the crucian carp's ability to plastically induce pupil enlargement should render fitness advantages, particularly under dim light conditions. Future studies should focus on the proximate mechanisms underlying this plasticity, and also include the density of retinal ganglion cells and image processing capacity in the brain, e.g. volume of the optic tectum, as visual perception is a product of a many-to-one mapping system (Wainwright, 2007). Such knowledge will further our mechanistic understanding of predator-induced phenotypic plasticity in the teleost pupil, as well as contribute to our general understanding of the evolution of the eye.

ACK N OWLED G EM ENTS
We thank Alexander Hegg and Ronja Zelmer for helping out with the experiments. Ethical permit for care and use of experimental animals were followed and provided by the Malmö/Lund Ethical Committee (M36-14).

DATA AVA I L A B I L I T Y S TAT E M E N T
Data are available from the Dryad Digital Repository https://doi. org/10.5061/dryad.0p2ng f1xr .