Functional effect of vaterite – the presence of an alternative crystalline structure in otoliths alters escape kinematics of the brown trout

The escape response of an organism is generally its last line of defence against a predator. In fishes, the fast-start escape response is the main locomotor behaviour observed in fish to evade predatory attacks and thereby increase their probability of survival (Walker et al., 2005). It consists of a contraction of the axial muscles that conduct to a brief high acceleration in a direction away from the threat (Weihs, 1973; Webb, 1978; Domenici and Blake, 1997). Such a startle response (referred to hereafter as ‘C-start’ in reference to the initial bending position of the body into a C-shape) is by far the best-studied behaviour pattern in terms of high-speed sensory motor control in fish (Karlsen et al., 2004). It has been studied from many perspectives, including kinematics, performance, physiology and neurobiology (Domenici and Blake, 1997; Korn and Faber, 2005). Under natural conditions the escape response is initiated by high-intensity acousticolateral, somatic or visual stimuli that activate giant reticulospinal neurons within the brainstem (the Mauthner cells, reviewed in Korn and Faber, 2005). Thus far, most of the studied C-start behaviours have been investigated in relation to extrinsic factors (i.e. both biotic and abiotic components) such as predator–prey or conspecific interactions, temperature, water velocity and viscosity, etc. In contrast, apart from ontogenetic aspects (i.e. body size), there has been surprisingly little consideration for inter-individual (i.e. between conspecifics) variability in mechanosensory systems and their possible consequences on escape performance. Such a lack of information mainly relies on the logistical difficulty in concurrently conducting appropriate behavioural experiments and investigating mechanosensory systems at the individual level. In an evolutionary context, phenotypic variation in sensory perception may be of interest as it is a component that selection might act upon (Langerhans and Reznick, 2011).

In the underwater environment, ‘sound’ propagates both as a pressure wave and as particle displacement, and therefore its detection should be interpreted as a multi-modal response, rather than as ‘hearing' (Higgs and Radford, 2013). The acousticolateralis system of fish is composed of two major structures (the inner ear and the lateral line) that offer a dual detection system to measure mechanical disturbances in the environment. They are both inertial motion detectors responsive to many of the same stimulus fields (Braun and Coombs, 2000; Mirjany et al., 2011). The inner ear is mainly responsible for balance and the detection of acoustic signals (Schulz-Mirbach et al., 2018a), whereas the lateral line detects water-borne vibration signals. The fish inner ear is composed of three connected semicircular canals and three otolithic end organs (Ladich and Schulz-Mirbach, 2016). The main constituent of otoliths is anhydrous calcium carbonate (CaCO₃) suspended in a protein matrix. Three crystalline polymorphs of calcium carbonate have been identified in teleost otoliths: calcite, aragonite and vaterite, in addition to calcium carbonate monohydrate (Gauldie, 1993; Campana, 1999). Sagittal otoliths, which are often the largest of the three types of otoliths, are normally composed of aragonite in many taxa. However, the inclusion of vaterite in abnormal or ‘crystallised’ sagittal otoliths has been reported in a number of marine and freshwater species from different environments (Gauldie, 1986; Strong et al., 1986; Gauldie et al., 1997; Bowen et al., 1999), with high prevalence in farmed fish (David et al., 1994; Sweeting et al., 2004; Reimer et al., 2016, 2017).

Due to their orthorhombic geometry, these ‘aberrant’ vateritic otoliths are significantly larger and less dense than their aragonite counterparts (Tomás and Geffen, 2003), leading to mass asymmetry in otoliths. Given the prominent importance of otolith mass symmetry in hearing (Lychakov and Rebane, 2005; Lychakov et al., 2006), the functional impact of vateritic sagittae on fish is therefore suspected to be considerable (Oxman et al., 2007; Reimer et al., 2016, 2017). There is currently strong theoretical and empirical (i.e. auditory brainstem response) evidence that vaterite deposition has a negative impact on auditory capacities in fishes (Oxman et al., 2007; Reimer et al., 2016). However, to date the functional response at a more integrated level (i.e. organismal level) has not been investigated. In particular, given the prominent role of the auditory system in escape trajectories in fishes (Popper and Carlson, 1998; Karlsen et al., 2004), whether change in the crystalline structure of otoliths is capable of altering fast-start escape in fish remains unexplored. Fishes are established animal models to study vestibular malfunctions. They possess no body weight-related proprioception to be used for maintenance of equilibrium and thus have to rely on their vestibular system for postural control (Anken et al., 2017). We therefore investigated the kinematics of brown trout during fast-start manoeuvres in relation to either the aragonitic or vateritic nature of otoliths. More specifically, we tested the hypothesis that individuals with vateritic otoliths have altered kinematic behaviour compared with those with fully aragonitic otoliths, both in term of performance and directionality. In addition, startle trials were conducted before and after chemical ablation of the lateral line in experimental conditions on the same individuals. This allowed focusing on the intrinsic effect of otolith-related hearing (i.e. in the absence of a functional lateral line) and examining to what extent the lateral line can compensate putatively altered escape responses in natural conditions.

For this experiment, a full-sibling group of juvenile brown trout (Salmo trutta Linnaeus 1758) spawned from anadromous parents from the Nivelle River, southwestern France (43°18′40″N, 1°31′53″W) were used. After fertilisation, eggs were transferred to fish hatchery facilities and incubated in natural temperature conditions (water derived from the neighbouring Lapitxuri stream, 43°17′00″N, 1°28′53″W; 10.5±0.9°C during the 2011–2012 winter). After completion of yolk sac absorption and the onset of exogenous feeding, all individuals were transferred and reared for 6 months (from the beginning of April to the end of September 2012) in one single experimental tank in the Lapitxuri experimental site (ECP, INRAE, 2018). During this period, all fishes were maintained under natural ambient temperature and ad libitum feeding conditions. As the time of completion for an escape response is size dependent, with larger fish performing faster/longer fast-starts (Webb, 1976; Domenici and Blake, 1993a), we randomly selected 60 individuals of equal size at the end of the growing period. Standard length (from snout to caudal peduncle) ranged from 58 to 69 mm (62±4 mm, mean±s.d.) and mass was 2.7±0.8 g (ranging from 2.2 to 3.5 g). At this stage, the presence/absence of vaterite in individuals remained unknown, with that information only being available post-dissection. Before the experiments, single individuals were kept undisturbed in a large tank (80×40×35 cm, length×width×height) filled to 30 cm depth with water (temperature 10.8±1°C) for at least 30 min in the dark.

All trials were conducted in an oblong acrylic experimental tank (Fig. 1A) with a laminar current flowing at 7±1 cm s⁻¹. In the set-up, two dividers (plastic net with a 3.0 mm mesh) were placed on both sides of the frontal part of the tank, delimiting a 300×130×60 mm (length×width×height) central area in which fish were positioned. The arena was illuminated with two 150 W spotlights, placed 0.6 m above the water surface. Individual fish were transferred into the experimental tank using a plastic container with a small volume of water to prevent any damage to their superficial neuromasts. After a few minutes, fish generally rested (i.e. stationary position) in a central position in the tank. Due to rheotaxis, the presence of a steady current maintained fish orientation upstream and parallel to the walls. After a minimum resting period of 3 min, individuals were subjected to an acoustic stimulus, inducing the fast-start escape behaviour. The startle stimulus consisted of a heavy plastic cylinder (340 g, 29 mm diameter, 45 mm height) released from 13 cm above the water surface and through a grey plastic pipe (35 mm outer diameter, positioned about 120 mm behind the fish, corresponding to about two body lengths away from the centre of mass of the fish), positioned 5 mm above the water surface to prevent the trout from seeing the stimulus before it made contact with the water surface. A transparent string was attached to the cylinder in order to prevent it from hitting the bottom of the tank. An additional acrylic divider was positioned above the aquarium (touching the water) preventing the distortions in the recorded video images due to the propagation of surface waves. As the distance between the acoustic/motion source and fish position may affect kinematics (Kimura and Kawabata, 2018), fishes were only stimulated when located about 11±2 cm from the stimulus. In addition, a fish was considered for stimulus only if its body centre of mass was at least 0.75 body lengths away from any wall and the anterior part of its body was oriented at an angle of less than 5 deg to the tank edge. Mathematical modelling studies indicate that the stimulus to the lateral line only changes within 0.25 body lengths of a wall (Hassan, 1992), so at three times this distance edge effects should be minimal on escape direction or performance. While we attempted to minimise the impact of these putative confounding variables (i.e. stimulus distance, wall distance and initial fish orientation) prior to data acquisition, we also a posterori tested for their individual effects (see below). Because fast-starts usually involve body bends and locomotion in the horizontal plane (Domenici and Blake, 1997), the swimming kinematics elicited by the stimulus were recorded using one camera that imaged the entire central area from above at 30 frames s^–1 (Sony XCD-X710, objective 6 mm).

To improve estimates of individual escape responses, all fishes were consecutively recorded four times, with a minimum resting period of 3 min allowed between replicates. If one replicate failed after the fish was startled (i.e. no escape response), another 3 min resting period was conducted before the onset of a new stimulus. On average, 95% of individuals responded to the first startle stimulus. In contrast, only half responded to the first stimulus of the fourth pseudoreplicate, with some individuals being stimulated several times before they escaped (see Fig. S1 for details). Between trials, an aerator and heater/cooler were placed in the experimental tank to maintain the water temperature (10.8±1°C) and oxygen level. In many experiments, fishes were not used more than once, to avoid habituation to the startle stimulus. To test for any possible effects that might be caused by learning, pseudoreplicates (four per individual) were primarily included in the statistical analysis (see below) as a fixed source of intra-individual variance (i.e. fixed effect term in mixed models). As pseudoreplicates treated as a fixed effect were all non-significant (P<0.01) and explained only a very small amount of variance (Tables S1 and S2), they were subsequently treated as a random effect (and simply referred to hereafter as ‘replicates’).

Both the fish inner ear and the lateral line are likely to play an integrative role in perceiving ʻacoustic' stimuli (Young Yan et al., 2010; Higgs and Radford, 2013), with the contributions of each being difficult to differentiate (Braun and Coombs, 2000; Webb et al., 2008; Mirjany et al., 2011). In order to avoid potential confounding effects related to the lateral line system and to specifically investigate the effect of otolith-related hearing, mechanoreceptors from the lateral lines were chemically inhibited. Fish were exposed to a solution of streptomycin (500 mg l⁻¹) and gentamicin (50 mg l⁻¹) for 3 h to block the mechanosensitivity of superficial and canal neuromasts without affecting the inner ear (Karlsen and Sand, 1987). Startle acoustic stimulation was first used on untreated fish (60 individuals, four replicates per fish, 3 min resting period between replicates) and then tested again using the same fish following chemical inhibition in the same conditions (60 individuals, four replicates per fish, 3 min resting period between replicates). For each individual, escape kinematics were thus investigated before and after inhibition of the entire lateral line. In this context, our experiment allowed us to investigate to what extent the lateral line can compensate putatively altered hearing capacities in natural conditions. At the end of the experiment, all individuals were euthanised by immersion in buffered benzocaine solution (preliminary anaesthesia at 4.0 ml l⁻¹ followed by euthanasia at 12.0 ml l⁻¹; 10–15°C). Experimental design and maintaining conditions were both approved by the National Ethic Committee for Fishes and Birds (CE73), with respect to the national recommendations. Dissection revealed the presence of individuals with either fully aragonitic otoliths or one-side-only vateritic otoliths, with the extent of vaterite deposition ranging from 8 to 72% (obtained from the orthogonal projection of opaque/aragonitic and transparent/vateritic zones).

Videos were analysed frame by frame from the first detectable movement following the stimulus for 480 ms (i.e. 16 images). All escape responses consisted of a first stage of ipsilateral activity, followed by contralateral activity (Domenici and Blake, 1997). We chose 480 ms as a fixed time in order to cover the entire fast-start behaviour, as well as some coverage of the fish having noticeably slowed after contralateral activity but before the onset of any potential subsequent swimming movement.

In each frame, the following points were manually digitised using ImageJ software (Rueden et al., 2017): the centre of mass (CM, located at the most anterior part of the visible dorsal fin), the most rostral tip of the snout, the base of the caudal fin and eyespots (Fig. 1B). In most studies, the analysis of distance-related parameters is based on the CM of the fish when stretched straight. This approximates the instantaneous CM, the point about which propulsive forces act (Webb, 1978; Domenici and Blake, 1997). As illustrated in Fig. 1C, the anterior part of the trout is not stretched straight between the snout and the CM. Therefore, we alternatively characterised the anterior midline of the body as the straight line passing from the tip of the snout to the mid-distance point between the eyes. In the absence of additional morphological landmarks, the straight line passing from the CM to the base of the caudal fin (i.e. tip of the caudal peduncle) is referred to hereafter as the posterior midline.

To simplify the kinematic analysis of a fish's motion, five swim metrics were calculated based on the above-mentioned reference points/lines (as shown in Fig. 1B,C). (1) A measure of effective swim distance (D) expressed as the straight line of the CMs between time steps t₀ and t_i. (2) A measure of two-dimensional trajectory efficiency (E) that is expressed as the ratio between D and the actual total travelled distance (R) within the same time period, expressed as the sum of the horizontal distances travelled by the CM between each successive frame (i.e. time steps t_i−1 and t_i, δd_i; Fig. 1C). For turning behaviour, overall bending/flexion was decomposed into sequences of anteroposterior motion as follows (Fig. 1B). (3) We measured the course of the head's angular changes by using the turning angle of the anterior midline between time steps t_i−1 and t_i (δΦ_i) and the associated cumulative turning angle between time steps t₀ and t_i (ΣΦ). (4) The tail bending was measured using the angular changes (deg) of the posterior midline between time steps t_i−1 and t_i (δθ_i), as well as the associated cumulative bending between time steps t₀ and t_i (Σθ). (5) Ultimately, the final escape angle was calculated using the position of the CM on the last frame, relative to the initial orientation of the anterior midline at rest, before stimulus onset. The unsigned escape angle (absolute value) was used to investigate the preferential escape directionality away from the stimulus. In addition, the signed escape angle (expressed positively when directed away from the vateritic side, and negatively when directed towards the vateritic side) was used to investigate whether individuals with vateritic otoliths escape preferentially towards the vateritic side due to biased directionality of sound localisation.

For illustrative purposes, outlines of the sharp silhouette of the fish viewed from above in successive frames were traced and digitised relative to fixed reference points (i.e. position of the CM before the first detectable movement following the stimulus onset).

Here, we tested the hypothesis that individuals with vateritic otoliths have altered kinematic behaviour compared with those with fully aragonitic ones. For this purpose, the trajectories of effective swim distance (D), cumulative turning angle (ΣΦ), cumulative tail bending (Σθ) and efficiency (E) over time (0–480 ms) were modelled using generalised additive mixed model (GAMM; Lin and Zhang, 1999) implemented using the mgcv package (version 1.8) in R version 3.2.2 (https://CRAN.R-project.org/package=mgcv). GAMM are semi-parametric models that utilise a data-driven approach to model a non-linear relationship between dependent and independent variables. The smoothing function models the potentially non-linear trajectory of kinematic measurements over time, with a possible optimisation based on a trade-off between model fit and model smoothness. Our main aim was not to directly compare kinematics in relation to chemical inhibition of the lateral line. Instead, we focused on differences between individuals having aragonitic or vateritic otoliths in different conditions (i.e. with or without a functioning lateral line) and models were fitted separately for each condition. ‘Time’ was used as the main independent variable, with ‘Vaterite’ (two groups, individuals with fully aragonitic otoliths and those with vateritic otoliths, as revealed after dissection) as a fixed effect, ‘Individuals’ (60 individuals) as a random effect that allows for the baseline value of kinematics (i.e. smoothness parameter) to vary for each fish, and ‘Replicates’ (four replicates per individual) as a random effect nested within individuals that can affect both slope and intercept of the individual smoothers. For a better interpretation, the Vaterite fixed effect was decomposed into a smoothing function (i.e. non-linear relationship) and a parametric term [i.e. intercept (constant) to be added to the global effect]. We also accounted for autocorrelation in the residuals due to the underfit of the model by including an AR1 model (i.e. first-order autoregressive model). Given the presence of numerous random effects and AR1 autocorrelation correction, the restricted maximum likelihood (REML) estimator was preferred for fitting models (Reiss and Ogden, 2009; Wood, 2011). Wald tests of the significance of each parametric and smooth term were performed. For visual inspection, the estimated trajectories were plotted with point-wise 95% confidence intervals. In addition, the difference between the two (non-linear) smooths comparing kinematic measurements of individuals with either aragonitic or vateritic otoliths were plotted with point-wise 95% confidence intervals [accounting for the parametric (intercept) term]. When the confidence band does not overlap with the x-axis, it indicates that difference is significantly different from zero.

Thus far, the presence of vaterite was considered as binary (fully aragonitic versus partially vateritic). To better link the effect of vaterite with swim metrics, we alternatively used the extent of vaterite deposition expressed as a percentage within GAMM. Time and ‘Vateritic extent’ were used as the main independent variables. The main non-linear interaction between Time and Vateritic extent was modelled using an anisotropic smoothing tensor within the GAMM. Additionally, Individuals (24 vateritic individuals) were treated as a random effect that allows for the baseline value of kinematics (i.e. smoothness parameter) to vary for each fish and Replicates (four replicates per individual) as a random effect nested within individuals that can affect both slope and intercept of the individual smoothers. To test for the effect of the Vateritic extent on escape trajectories, we separately reported its own significance (Vateritic extent treated as a main effect), as well as its change over time (i.e. Vateritic extent and Time interaction).

The signed angles were categorised using either positive or negative values and tested for homogeneity using a chi-squared test (test for preferential escape towards the vateritic side). In the absence of significance, differences in unsigned terminal (i.e. 480 ms) escape angles were tested using the Watson–Williams test (test for homogeneity of means between several samples of circular data). This tests for preferential escape directionality away from the stimulus.

A posteriori dissection revealed that among the 60 fish used in the experiment, 24 individuals had vateritic otoliths and the remaining 36 were fully aragonitic. The comparison between individuals with aragonitic versus vateritic otoliths was not biased by the initial position of the fish (i.e. initial body angle relative to the flow direction and distance from the edge) or by its distance to acoustic stimuli in any of the trials (insignificant correlation between these initials parameters and kinematic metrics established at any fixed time, rho<0.1 for both).

The escape kinematics globally exhibited consistently distinct characteristics between individuals with fully aragonitic or vateritic otoliths, as revealed by contrasting time-dependent parameters (i.e. effective travelled distance, turning angle, tail bending and trajectory efficiency over time). According to the GAMM modelling, both smooth and parametric terms were significantly different (for details, see Table 1 and Table S3), indicating that all kinematic parameters differed not only on average but also in the shape of their functional response over time. More specifically, escape trajectory is a two-stage motor process: fish first respond with a unilateral contraction of their axial muscle (stage 1) that results in bending their body into a C-shape directed away from the threat, which is followed by a contralateral contraction (stage 2), thus pushing off water with the full broadside of its body (Domenici and Blake, 1997), and possibly continuing by one or more tailbeats. The end of stage 1 was therefore defined as the reversal of the turning direction of the head. In this context, escape trajectories in individuals having vateritic otoliths imply larger overall body flexion during stage 1. Data suggest that postural curvature similarly affects the head and the tail (as expressed by turning angle and tail bending, respectively; Fig. 2A,C). Due to this pronounced initial curvation, the anterior body midline of individuals having vateritic otoliths is brought to a position almost perpendicular to the original body axis, and the CM accordingly undergoes a slight displacement in this orthogonal direction (Fig. 3). As the body axis was aligned with the flow direction at the end of the burst in all individuals, the exaggerated ipsilateral postural curvature observed in individuals having vateritic otoliths is associated with a much more pronounced contralateral contraction (Fig. 3). This therefore requires a longer initial bending phase and results in a delayed forward acceleration of individuals with vateritic otoliths. In the absence of any further compensatory mechanism, this primary combination of latency and orthogonal displacement leads to a decrease in efficiency that propagates over time, until the end of the escape response (Fig. 4A). As a corollary, while individuals with vateritic otoliths escaped with a slightly lower effective distance at each given time, individuals with fully aragonitic otoliths escaped further with better efficiency (Figs 4A and 5A). Overall, the mean effective escape distance after 480 ms was reduced by about 15% in individuals with vateritic otoliths compared with those with aragonitic ones (unilateral Wilcoxon–Mann–Whitney test with continuity correction for comparison of means, W=5778, P=0.015; see density curves in Fig. 5A).

In the presence of pharmacologically inhibited neuromasts, fish still exhibited a rheotaxic response (probably due to visual/proprioceptive cues) and still escaped (suggesting that behavioural responses were not fully dependent on a functional lateral line). However, blocking superficial and canal neuromasts from the lateral line with a chemical treatment altered escape kinematics in our experimental design. Compared with control animals, both aragonitic and vateritic trouts with a blocked lateral line depicted slightly more curved frontal (Fig. 2A,B) and caudal (Fig. 2C,D) parts and escaped at closer range (Fig. 5A,B) with roughly equal efficiency (Fig. 4A,B). In addition, they exhibited larger escape angles with the centre body axis (Fig. 6A,B). Besides these aspects, differences between individuals with either aragonitic or vateritic otoliths were more pronounced when mechanoreceptors from the lateral lines were chemically ablated (i.e. ‘acoustic' stimuli being perceived only by the inner ear). In particular, trajectory efficiency was significantly diminished in individuals with vateritic otoliths compared with those possessing fully aragonitic otoliths (Fig. 7). Overall, mean effective escape distance after 480 ms was reduced by about 25% in individuals with vateritic otoliths compared with those with aragonitic ones (unilateral Wilcoxon–Mann–Whitney test with continuity correction for comparison of means, W=4236, P=1.91×10⁻⁷; see density curves in Fig. 5B). As previously established, differences in performance arise early during stage 1, propagate throughout the whole response and consist of an exaggerated bending of both the tail and the head with respect to the original body axis.

Thus far, the plots of predicted GAMM for individuals with vateritic and aragonitic otoliths (Figs 2, 4 and 5) is inadequate for significance testing, and should only be used for illustrative purposes. Alternatively, one can plot the difference smooth itself with a confidence interval and check if the confidence interval includes 0 at different points. All comparisons are given in Fig. 7, both in presence and absence of a functional lateral line.

In addition, considering the presence of vaterite as a continuous trait (percentage) emphasises the intrinsic functional role of vateritic otoliths in escape trajectories (Table 2; Fig. 8). Partial effects predicted by GAMM highlight that while the extent of vaterite deposition increases, cumulative turning and bending angles increase (Fig. 8A,B). Simultaneously, efficiency and effective escape distance decrease (Fig. 8C,D). These differences are more pronounced in the absence of a functional lateral line, although with a larger confidence interval. This is especially clear for the efficiency as this metric summarises all other trajectory characteristics.

The signed escape angle (expressed positively when directed away from the vateritic side, and negatively when directed towards the vateritic side) exhibited no bias towards the vateritic side (χ²=1.04, d.f.=1, P=0.307 for functional lateral line, and χ²=2.04, d.f.=1, P=0.153 for lateral line chemically ablated). In addition, escape response in individuals with fully aragonitic otoliths also exhibited no bias towards a particular side (χ²=1.00, d.f.=1, P=0.317 for functional lateral line, and χ²=2.25, d.f.=1, P=0.133 for lateral line chemically ablated), highlighting the absence of vortex, or non-laminar current flow in the experimental tank that would have resulted in a preferential escape side. In addition, while escape angles (defined as the angle between the body axis of the fish at rest and the position of the CM at the end of the record) cover the entire 0–60 deg range on either side of the fish, the large majority of unsigned angles is preferentially located at 10–30 and 30–45 deg away from the stimulus, respectively, before and after disruption of neuromast function (Fig. 6). In either condition, the presence of vaterite did not alter the directionality of the escape trajectories relative to the stimulus direction, as revealed by highly similar unsigned escape angle (Watson–Williams test for homogeneity of means, mean angle_aragonitic=32.1 deg, mean angle_vateritic=27.8 deg, F_1,238=1.36, P=0.243 for functional lateral line, and mean angle_aragonitic=37.9 deg, mean angle_vateritic=33.3 deg, F_1,238=1.95, P=0.164 for lateral line chemically ablated; Fig. 6).

Our results highlight that the presence of an alternative crystalline structure in otoliths is associated with altered escape kinematics in the brown trout. While the exact underlying mechanism cannot be fully understood without a more thorough understanding of the factors influencing escape trajectories, it is likely that behavioural differences are largely caused by the presence of vaterite. However, the precise cause–effect relationship is still unclear and one may not exclude the fact that individuals constitutively exhibiting altered kinematics are also more prone to developing vateritic otoliths. In this scenario, the presence of vaterite would appear concomitantly with altered kinematics in the absence of a direct functional link. Yet the close link between the individual extent of vaterite and associated decrease in escape performances (Fig. 8) strongly suggests a causal relationship.

The deposition of calcium carbonate in its crystalline vateritic polymorph has a significant effect on the escape kinematics of fish. The most curved postures observed during stage 1 are generally thought to increase subsequent locomotor performance as a higher amplitude of the tail should provide higher thrust (Turesson et al., 2009). It is, however, not the case in our results with the exaggerated ipsilateral contraction observed in individuals with vateritic otoliths being followed by a contralateral postural curvature. The associated temporary displacement of the CM in a direction orthogonal to the final escape angle acts as a latency (longer rotational phase) that propagates through the entire sequence and reduces distance–time performances (i.e. escaping at closer distance at any given time with lower efficiency). Lower distance–time performances are generally associated with higher angles of turn as the time of completion for an escape response is linearly related to turning angle (Domenici and Batty, 1994; Nair et al., 2015). This is especially significant for ‘double-bend’ starts that are naturally observed in several fish species (Domenici and Blake, 1991; Kasapi et al., 1993; Domenici and Blake, 1997). In addition, long latency response may enable fish to be more accurate in discerning the direction of the threat (Domenici and Batty, 1994). This, however, was not the case in our study, given the highly similar escape angles between individuals with aragonitic and vateritic otoliths.

Although C-starts are usually mediated by the Mauthner neurons and associated networks (Eaton and Emberley, 1991; Eaton et al., 1995; Guzik et al., 1999; Korn and Faber, 2005), it is not clear what neuromuscular component is altered by the presence of vaterite. It is, however, likely that the peculiar kinematics observed in individuals with vateritic otoliths rely on defects in mechanosensory systems, altering auditory function (Oxman et al., 2007). Vaterite is less dense than aragonite, and is often deposited in an irregular pattern on the otolith, leading to abnormal shape (Zhang et al., 1995; Bowen et al., 1999). This combination of altered density and morphology may impair the sagitta's ability to interface with the sensory hair cells on the epithelium. The sulcus is a depression on the medial surface of the otolith located where sensory tissue comes into contact with the otolith. Experiments conducted under altered gravity suggest a strong regulation of otolith size in order to maintain an equilibrium in the force exerted on underlying hair cells (Anken et al., 1998, 2000, 2002). These data lead us to consider that otolith mass affects hair cell excitation by gravity. Any change in otolith mass or size due to vaterite would therefore result in an altered perception or change in the range of accelerations (i.e. particle motion of the sound wave) to which the otolith organ is sensitive (Lychakov and Rebane, 2000, 2005). In addition, saccular epithelium was suspected to be altered (i.e. fewer hair bundles) in the presence of vateritic sagittae, but investigations remained inconclusive (Oxman et al., 2007; Brown et al., 2013). Altogether, this combination is known to erode hearing abilities in a considerable manner (Oxman et al., 2007; Reimer et al., 2016, 2017). For example, Oxman et al. (2007) described in Chinook salmon (Oncorhynchus tshawytscha) a significant loss of hearing sensitivity in vateritic fishes by using an auditory brainstem response technique. Similarly, Reimer et al. (2016) estimated a 28–50% loss of otolith functionality in individuals with vateritic otoliths using a mechanistic model of otolith oscillation.

The regulation of otolith formation is probably controlled by a number of genetic and neuroendocrine factors, and the perturbation of one or more of these factors may cause the shift to vateritic otolith formation (Söllner et al., 2003; Tomás and Geffen, 2003; Payan et al., 2004). In addition, it is worth considering if the mechanisms that cause vateritic otoliths might also contribute to deformation of neuromasts in the lateral line as both alterations may appear concomitantly (Brown et al., 2013). Of particular interest, several authors have reported that large numbers of vateritic otoliths (up to 50–60%) are frequently observed in farmed fish, compared with wild populations (David et al., 1994; Bowen et al., 1999; Sweeting et al., 2004; Reimer et al., 2016, 2017). As the kinematics and performance of fish during fast-start manoeuvres may determine the outcome of predator–prey interactions in terms of survival (Domenici and Blake, 1997; Walker et al., 2005; Walker and McCormick, 2009; Marras et al., 2011), the presence of vaterite may be of concern for the effectiveness of large restocking programmes based on captive-bred fish that are conducted worldwide (Bell et al., 2006; Cooke and Cowx, 2006; Bell et al., 2008; Brown et al., 2013; Taylor et al., 2017). However, it remains unclear if diminished performance observed early in experimental facilities can subsequently alter an effective evasive manoeuvre after release in the wild. Several authors have highlighted that kinematics patterns change considerably through ontogeny (Webb, 1976; Domenici and Blake, 1993a; Hale, 1996, 1999). Authors have pointed out both physiological (i.e. neuromast proliferation; Higgs and Fuiman, 1996) and mechanical constraints (i.e. limited flexibility at the CM in large fish; Domenici and Blake, 1997). In addition, larval fish have the potential to deviate from the adult models of neuromechanical control (Nair et al., 2015). As such, the kinematic consequences of vateritic otoliths in adults remain unexplored and require further investigation. Similarly, the direct comparison between wild-reared and hatchery fish would constitute a stimulating practical direction for future research.

The ability to escape predators depends not only on locomotor performance but also on a number of variables that are related mainly to the behavioural decisions of fish (i.e. including responsiveness, escape latency, reaction distance and directionality; Walker et al., 2005; Turesson et al., 2009). In particular, a key feature of the fast-start escape response is the consistent orientation of the response away from the threatening stimulus source (Eaton et al., 1995; Karlsen et al., 2004). Eaton and Emberley (1991) demonstrated an accurate inverse relationship between the angles of acoustic stimuli and the angular components of the response movements, suggesting that the fish measures the sound source angle (Domenici and Blake, 1993b). In fishes, both the mechanosensory lateral line organ and the inner ear are inertial motion detectors responsive to many of the same stimulus fields (Braun and Coombs, 2000; Mirjany et al., 2011). By selectively removing the lateral line, several authors have highlighted that this system is more likely to encode directionality in the escape response (Mirjany et al., 2011). However, the inner ears may also be used to determine the three-dimensional directionality of an incident sound wave to a lesser extent (Sand, 2002; Coffin et al., 2014). It has been proposed that fish localise sounds by comparing the phase of the non-directional sound pressure generated by the swim bladder and the directional particle motion (acceleration) resulting from the displacement of otoliths (Schuijf and Buwalda, 1975; Popper and Fay, 1993; Eaton et al., 1995; Guzik et al., 1999). In this context, it is noticeable that crystallised (i.e. vateritic) sagittae are less dense than their aragonite counterparts (density of 2.95 and 2.54 g cm^–3, respectively, for bioaragonite and biovaterite; Tomás and Geffen, 2003), leading to mass asymmetry in otoliths that can change the phase difference between these components. This can alter sound localisation capability and consequently may change directionality of the escape response.

An oscillatory model of fish hearing predicts that if the sound direction is parallel to the longitudinal axis of the fish and the mass of the right otolith is heavier, the pattern of otolith displacements is identical in shape and amplitude to the response pattern in a fish without otolith mass asymmetry but with the sound direction biased to the right side (Lychakov and Rebane, 2005; Lychakov et al., 2006). In other words, when the sound direction is parallel to the longitudinal axis of the fish (as was the case in the present study), it will appear to a fish that the sound source is situated at an angle to its longitudinal axis. We may therefore expect an altered directional decision-making process during an escape in individuals with vateritic otoliths. Contrary to these theoretical expectations, our data provide no support for this hypothesis in the absence of a functional lateral line, with fish still escaping preferentially at the same angle away from the stimulus, on either side with no bias of the signed escape angle towards the vateritic side. This may be partly explained by the extreme decrease in left–right discrimination when the stimulus is more in line with the longitudinal body axis (Domenici and Blake, 1997), as was the case in the current experimental design. Thus, we expect intrinsic reduced precision in the escape directionality regardless of the crystalline structure in otoliths. In the present situation, the apparent deflection angle between the true and sham sound directions is theoretically only a few degrees (maximum of about 5 deg for an average otolith mass of 200 µg; for details, see Lychakov and Rebane, 2005). Our experiment therefore has limited statistical power, precluding highlighting such subtle effects. It would be interesting to investigate whether and to what extent stimuli from different directions affect the directional change in escaping in individuals with vateritic otoliths. Such experimental work is particularly important as recent in situ investigations of sound-induced otolith motion yielded partially conflicting results compared with an oscillatory-based model, raising doubts about the validity of theoretical expectations (Schulz-Mirbach et al., 2018b).

Although data revealed pervasive significant effects after pharmacological ablation of neuromasts from the lateral line, they also emphasise that otolith-related behavioural alteration is partly compensated for in natural conditions. As already stated, mechanosensation from both the lateral line organ and the inner ear can detect displacement components of sound stimuli (Braun and Coombs, 2000; Mirjany et al., 2011; Higgs and Radford, 2013). It is likely that in nature fish use inputs from both auditory and lateral line systems either simultaneously or in series to interpret ‘sound’ stimuli and make appropriate behavioural decisions (Braun and Coombs, 2000; Webb et al., 2008; Higgs and Radford, 2013). As such, there is a degree of functional redundancy between these two modalities. Incongruence between signals recovered from these two inertial motion detectors could have resulted in more pronounced behavioural differences between individuals having aragonitic or vateritic otoliths. That, however, was not the case. While it is possible that some neuromasts survived streptomycin/gentamicin treatment (Brown et al., 2011), it is clear from the behavioural results that a reduction in neuromast inputs significantly enhances kinematical differences between individuals having aragonitic or vateritic otoliths. This emphasises that the crystalline structure of the otolith has an intrinsic functional effect on fish that is partially compensated for by the presence of a functional lateral line.

In the underwater environment, sound propagates both as a pressure wave and as particle displacement (Higgs and Radford, 2013), with particle displacement dominating close to the source (Higgs and Radford, 2013). In our experimental design, it was not possible to separate both components of sound when a heavy plastic cylinder striking the water surface was used to test hearing capacities. In the nearfield, both inner ear and neuromast hair cells are potentially stimulated by the particle motion component of sound sources, with the relative contribution of each changing as the distance from the sound source increases (Rogers and Cox, 1988; Montgomery et al., 2006; Higgs and Radford, 2013). The lateral line operates over a shorter distance range than the inner ear, although with a much greater spatial resolution (Montgomery et al., 1995; Braun and Coombs, 2000; Bleckmann and Zelick, 2009). Experiments were all conducted in enclosed acoustic environments with the fish well within the nearfield and with the sound propagation being notoriously complex (Higgs and Radford, 2013). As a consequence, direct extrapolation of our results to open water, where sound linearly propagates as a pressure wave, would be speculative. However, we may expect amplified behavioural differences between individuals having aragonitic and vateritic otoliths in such environments as long-distance acoustic stimuli would mainly involve or rely on the inner ear, with limited possible compensation from the lateral line.

Moreover, the escape response is an integrative process usually controlled by the Mauthner cells, a pair of large reticulospinal neurons within the brainstem that receive and integrate various sensory inputs, including visual and mechanoacoustic elements (Eaton et al., 2001; Korn and Faber, 2005; Higgs and Radford, 2013). Behavioural responses may thus vary according to complex three-dimensional hydrodynamic environmental conditions (Liao, 2007; Higham et al., 2015) and should be investigated in a context-dependent manner (Domenici et al., 2011). Changes in escape performances for individuals with vateritic otoliths, if any, thus remain to be investigated in more realistic and natural conditions, even in the wild.

Fish mainly rely on motion and vibration detection to establish orientation, maintain equilibrium, and interpret their surroundings, including obstacle detection, congener identification (e.g. competition, reproduction, school cohesion), prey localisation and predator avoidance (Pitcher, 1979; Popper and Coombs, 1980; Fay and Popper, 2000; Popper et al., 2003). In this context, the possible defects in mechanosensory systems may have various (subtle) consequences on the ecology of fish (Slabbekoorn et al., 2010). Surprisingly, all these ecological and functional aspects have already been investigated in conjunction with defects in the lateral line system. In contrast, the functional implications of a defect in the otolith-related hearing function due to vaterite have remained largely unexplored, although hypothesised, particularly in relation to fish behaviour [but see for example Anken et al. (2017), investigating space motion sickness under diminished gravity due to otolith mass asymmetry]. This is, to the best of our knowledge, the first demonstration that the crystalline structure of the otolith alters the escape kinematics of fish in a way that can possibly increase their predation. However, further behavioural studies coupled with neurological and physiological assays are required for a better in-depth understanding of the functional and ecological significance of the presence of abnormal crystalline structures in otoliths.

The authors are grateful to Stéphane Glise for help in maintaining fish for the duration of the experiment, to Jacques Rives for extracting/photographing otoliths, and to two anonymous reviewers who made constructive suggestions on the first version of the manuscript.