Development of sound production in Danionella cerebrum Free

Acoustic communication can be found throughout the animal kingdom in rich spectral and temporal diversity. Although the mechanisms of sound production are manifold, from bone stridulation to larynx vibrations (Suthers et al., 2016), a common theme is that sounds are associated with reproductive and territorial behaviour.

In sound producing species with parental care, acoustic signals are typically produced already in early developmental stages, for instance as retrieval calls upon maternal separation, i.e. in bats (Bohn et al., 2007), squirrel monkeys (Symmes and Biben, 1985), rats (Brudzynski et al., 1999) and seals (Sauvé et al., 2015), or as food calls of nestlings (Briskie et al., 1999). Vocalizations change throughout the lifespan of an animal, especially around the time of sexual maturation. A well-known example is the change in pitch in human males during puberty. In many vocal learner species, individuals acquire a new vocal repertoire upon sexual maturation, as prominently studied in songbirds (Nottebohm, 1970). The trajectory of sound development of vocal learners can vary. For instance, in the greater sac-winged bat, pups already start vocal imitation of male territorial song (Knörnschild et al., 2010), whereas pups of neotropical singing mice are highly vocal right after birth, but stop vocalising before weaning and then produce their mature advertisement song de novo (Campbell et al., 2014).

In soniferous egg-laying species without parental care, acoustic communication is present in adults and tends to commence in juvenile stages. For instance, tadpoles do not emit sounds, but vocalisations are found in juvenile spadefoot toads (Ten Hagen et al., 2016). In several sound producing species of the Sciaenidae, the sound generating muscles only form in juvenile fish in parallel with gonad development (Hill et al., 1987). In contrast, it has been reported that coral reef grey snapper already produce nocturnal sounds as pre-settlement larvae with a body length of a mere centimetre (Staaterman et al., 2014). This suggests that within the teleostean fishes, there is variability in when and how the sound production systems develop in an animal's lifetime. In addition, there is limited information to date linking the morphological ontogenesis of the sound producing organs to sound production and the resulting sound characteristics across development.

Here, we investigated the interplay between morphological ontogenesis and the development of sound production in the transparent miniature teleost Danionella cerebrum. This fish has an adult body length of only ∼12 mm and the smallest known vertebrate brain (Britz et al., 2021; Penalva et al., 2018 preprint; Schulze et al., 2018). Its small size, together with its life-long transparency, has made D. cerebrum a useful model system for neuroscience, suitable for whole-brain optical measures of neuronal activity (Schulze et al., 2018; Kadobianskyi et al., 2019; Hoffmann et al., 2023; Lam, 2022; Zada et al., 2024; Lee and Briggman, 2023). Adult fish have a male-specific trait of acoustic signalling (Schulze et al., 2018; Cook et al., 2024; Vasconcelos et al., 2024). Sounds can be observed in social contexts, including during courtship and male aggression (Schulze et al., 2018; Vasconcelos et al., 2024). They are composed of discrete pulses with a broad frequency spectrum (Cook et al., 2024; Vasconcelos et al., 2024), hereafter referred to as clicks. These clicks are produced by the contraction of a drumming muscle, causing a piece of cartilage to snap against and compress the anterior swim bladder (Cook et al., 2024). Sequences of clicks appear at repetition rates of either ∼60 or ∼120 Hz, caused by unilateral or alternating bilateral muscle contractions (Cook et al., 2024). The highly specialised drumming apparatus is only present in males and is absent in females (Britz et al., 2021). To date, it is unknown when D. cerebrum develops its sexually dimorphic sound production apparatus and at which developmental stage acoustic signalling begins. We addressed these questions through a combination of longitudinal sound recordings and morphological comparisons.

All animal experiments conformed to Berlin state, German federal and European Union animal welfare regulations. Danionella cerebrum Britz, Conway & Rüber 2021 were housed in water circulated Tecniplast aquaria with artificial fish water (Instant Ocean at a conductivity of 350 µS cm⁻¹, adjusted to a pH of 7.3 using NaHCO₃) at 27°C and a 14 h:10 h light:dark cycle, with lights on from 08:00 until 22:00 h preceded by a 30 min dimming period. For all behavioural experiments, wild-type (WT) fish were used. For morphological stainings, a tyrosinase knock-out line (tyr−/−) was used to remove melanin pigmentation for increased transparency to see bone and cartilage structures.

Fertilised eggs from communal breeding in the regular home tanks were collected and kept in 10 cm Petri dishes containing embryo water (600 mg l⁻¹ Instant Ocean, 150 mg l⁻¹ NaHCO₃, 50 mg l⁻¹ CaCl₂ 2H₂O, 0.5 mg l⁻¹ Methylene Blue) at a density of <20 eggs per dish. At 5 days post fertilisation (dpf), larvae were transferred to 1 litre plastic tanks (Tecniplast breeding tank) with A2000 water (2 g l⁻¹ Instant Ocean; aquarium system, 150 mg l⁻¹ NaHCO₃, pH 7.4, conductivity 1.5 mS cm⁻¹) and L-type rotifers (Brachionus plicatilis, planktovie.biz). For regular facility rearing, larvae were transferred to the water circulation system at 10 dpf and fed twice daily with Artemia, as well as dry food (Gemma micro 75). For rearing in experimental tanks, larvae were transferred at 3 weeks post fertilisation (wpf) and fed once daily with Artemia. Water was either circulated by a pump or manually exchanged twice a week in the small tanks. All recording rooms were temperature controlled at 27°C.

For the single recording of a large group (Fig. 1A), 100 larvae at 3 wpf were transferred to a stand-alone glass aquarium (30×60×30 cm) filled to a height of 10 cm with artificial fish water with a circulating pump. Three times per week, 1 litre of tank water was replaced with fresh artificial fish water. Water temperature was maintained at 27°C with three aquarium heater thermostats (Sera). Once a week in the morning between 09:00 and 11:00 h, a hydrophone (H3x, Aquarian Audio) was inserted into the tank and 1 h sound recordings were acquired with a datalogger (Zoom field recorder, Fn8) at a sampling rate of 48 kHz, while the circulating pump was switched off.

For the multiple group recordings (Fig. 1B–J), 10 larvae at 3 wpf were kept inside 15×15×10 cm acrylic tanks filled to a height of 8 cm with artificial fish water (at 27°C). This led to a higher fish density than in the large aquarium or the fish facility, but it was a trade-off between recording quality and fish number. Because a higher fish density can lead to developmental delays, we measured body size in addition to biological age. To obtain length measurements without disturbing the recordings, we raised five additional groups of 10 fish in equivalent acrylic tanks, kept in the same recording room and treated under the same conditions apart from hydrophone recordings. Five of these sample fish were collected per week (one fish per tank), euthanised and photographed under a brightfield microscope (Olympus MVX-10 with a MV PLAPO 1X objective, connected to a FLIR usb colour camera). Standard length (SL) measurements were manually performed post hoc on the collected images, as measured from snout to the base of the tail, excluding the tail fins. Owing to the sampling, the fish density in the extra tanks was decreasing each week, which could have led to inflated length measurements. We considered this factor to be negligible, in comparison with the length variability at each time point and because the values for weeks 11 to 14 are within the known range of adult male and female D. cerebrum.

To avoid disturbance of the fish prior to sound recordings, a hydrophone (AS-1, Aquarian Audio; linear frequency range from 1 Hz to 100 kHz) was placed inside each tank the evening prior to a 2 h recording from 09:00 to 11:00 h the following morning, a time of the day when D. cerebrum have previously been shown to be sonically active (Schulze et al., 2018; Vasconcelos et al., 2024). Juvenile fish follow a similar circadian pattern of sound production during the light hours, with activity peaks during the first half of the day (Fig. S1). Recordings were repeated twice a week over the course of 11 weeks, to record ages 4–15 wpf. Sound recordings were obtained with a sampling rate of 51.2 kHz via a National Instruments data acquisition card (NI-9231) controlled through a custom-written Python script. Each AS-1 hydrophone was amplified with a PA-4 hydrophone preamplifier at an inverted wiring state (Aquarian Audio), powered by a 12 V battery.

Three age-matched males were placed in a 5×5×5 cm box made out of transparent acrylic and walls of blue cling film. This box was placed in a holder in the centre of a 27×27×30 cm glass aquarium filled to ca. 27 cm. Five AS-1 hydrophones were positioned around the four vertical walls and the bottom of the inner box, with a fixed distance of 3.5 cm to the centre of the box. Sound recordings were acquired for several hours at a sampling rate of 102.4 kHz using a National Instruments data acquisition card (NI-9250).

Clicks that were detected on at least four hydrophones at coinciding or adjacent samples were considered centre clicks, produced by fish at the centre of the box and hence an equal distance of ca. 3.5 cm. To control for the different amount of recorded centre clicks per group, 100 centre clicks were randomly selected and out of those the 10 loudest were determined. For each of these 10 clicks per age, the sound trace from the channel with maximum amplitude was used for further analysis. We used the average of the 10 clicks for statistical comparison of sound amplitude and peak-to-peak time across age, to avoid pseudo-replication and thus inflated P-values. Regardless of age, not all groups of fish produced sounds in this configuration. We only considered groups with at least 100 detected centre clicks for the analysis.

The fish used in this experiment were raised in the facility and had a faster growth rate than the repeated-measures experimental fish. Therefore, the colour code for fish maturation (Fig. 2C–E; Fig. S2) is estimated based on swim bladder height as measured after the sound recordings and expressed as the mean between the three fish of each group. The matching between biological ages, SL and swim bladder height (SB height) were: 5-week-old fish, SL=9.3±0.4 mm, SB height=0.68±0.04 mm; 6-week-old fish, SL=9.7±0.6 mm, SB height=0.75±0.06 mm; 8 week old fish, SL=10.5±0.2 mm, SB height=0.85±0.03 mm and adult (older than 3 months), SL=10.5±0.6 mm, SB height=0.92±0.07 mm. Because we did not assign clicks to individual fish, low within-group variation of SB height is required to meaningfully relate SB height to click parameters. Therefore, we excluded one group of juvenile fish with an unusually large variation in SB height.

Clicks were detected with a custom-written Python code. Waveform traces were bandpass-filtered between 3 and 20 kHz for repeated-measures recordings and 1 and 20 kHz for the close-range click recordings. Peaks were detected using the waveform envelope and a refractory period of 5 ms between peak times. Filtering parameters and detection thresholds were always kept constant for all recorded ages within one experiment. Bursts were defined as sequences of clicks appearing at less than 25 ms between two clicks.

Tyr–/– specimens of different ages were euthanised in ice water and fixed overnight in 4% paraformaldehyde in 1× PBS at 4°C. After brief washing in 1× PBS, samples were stained for 24 h with a double staining solution of 0.001% Alcian Blue (Roth) and 0.005% Alizarin Red S (Sigma-Aldrich) with 200 mmol l⁻¹ MgCl₂×6 H₂O in 70% EtOH (Walker and Kimmel, 2007). After two brief washes in pure H₂O, samples were kept in 20% glycerol with 0.25% KOH for 2 h and 50% glycerol with 0.25% KOH overnight. Thereafter, samples were kept in a commercial refractive index matching solution (RI=1.465, EasyIndex Life Canvas) for 2–3 days and imaged under a brightfield illuminated microscope (Olympus MVX-10). All steps after fixation were performed at room temperature. SL measurements were taken post mortem prior to fixation to account for possible shrinkage owing to sample dehydration.

To probe at what developmental stage D. cerebrum first emit sounds, we raised a group of 100 fish in a stand-alone community aquarium and performed underwater sound recordings once a week. The first time point at which click sounds were clearly distinguishable from background noise was 6 wpf. From 6 to 8 wpf, we observed an increase in the abundance as well as the amplitude of clicks (example sound traces shown in Fig. 1A).

To examine sound production in more detail across ages, we next raised six groups of 10 fish in 15×15×10 cm tanks, equipped with one hydrophone each, and recorded twice a week over the course of 11 weeks starting at 4 wpf. To accurately measure the growth rate without disturbing the sound recording tanks, we raised additional individuals under identical conditions and measured the SL (distance from snout to tail-base). Fish grew over the course of 3 months until they reached a SL of 11.9±1 mm (N=5 per age; Fig. 1B). These measurements were drawn from a random pool of a mixed-sex fish population. Females are generally larger than males at 15 wpf (11.8±0.7 and 10.8±0.6 mm, respectively, two-sided t-test: t=5.3, P<0.001, N=37 females, N=23 males). Growth in body size was accompanied by gross morphological changes, including a change in the shape of the swim bladder complex (Fig. 1C).

Similar to the communal aquarium recording, the number of recorded clicks increased between 6 and 9 wpf (Fig. 1D), corresponding to 8.1±1.3 mm SL at 6 weeks and 9.1±0.8 mm SL at 9 weeks (Fig. 1B). The exact age when groups started producing clicks varied between 6.5 and 8 weeks in the different groups (see grey lines in Fig. 1D). The average age with more than 20 recorded clicks per group was 7±0.5 weeks.

Adult clicks appear with a repetition rate of either ∼60 or ∼120 Hz (Schulze et al., 2018; Vasconcelos et al., 2024), corresponding to repeated unilateral and alternating bilateral contractions of the drumming muscles (Cook et al., 2024). Both click repetition rates were present in juveniles from the earliest sound-producing age onwards (Fig. 1E), and the peak of the bimodal distribution did not change with age (Fig. 1F). Groups of repeated clicks, hereafter referred to as bursts, tend to occur with a click repetition rate of either 60 or 120 Hz (example bursts shown in Fig. 1G). Similar to the total number of clicks (Fig. 1D), the number of bursts increased between 6 and 9 wpf (Fig. 1H). In total, bursts with a single click repetition rate (i.e. either 60 or 120 Hz) make up approximately 80% of all detected bursts throughout all ages tested. At younger ages, D. cerebrum tend to produce more 60 Hz bursts, while at later stages bursts of either click repetition rate occur at a similar scale (Fig. 1I). Finally, we measured the length of each burst as the number of clicks in the sequence (Fig. 1J). The maximum number of clicks observed in a burst reached up to several hundred for 120 Hz bursts in older fish, while for 60 Hz bursts it did not surpass 10.

In the communal aquarium recordings (Fig. 1A), as well as the group recordings (Fig. 2A), the amplitude of the clicks increased with age. Because clicks are generated by compressing the anterior swim bladder (Cook et al., 2024) and the swim bladder size increases with age (Fig. 2B), we next asked how click features change as a function of swim bladder size. Because in the previous sound recordings the distance of the sound source to the hydrophone was unknown, the sound attenuation and thus the actual sound volume could not be reliably established.

We therefore performed multi-hydrophone recordings of clicks at a fixed distance at four developmental time points (see Materials and Methods for details). These click recordings allowed us to compare the click profile, which was stereotyped within each developmental group (Fig. 2C). Although the shape of the pulse was similar across stages, with its peaks aligned in time, click amplitude was positively correlated with SB height (Pearson r=0.99, P=0.013, N=4; one data point per group; Fig. 2D), and reached up to 140 SPL (dB re. 1 µPa), matching previous reports of D. cerebrum sound pressure levels (Cook et al., 2024). In contrast, the peak-to-peak time of the main click pulse was not correlated with SB height (Pearson r=0.20, P=0.80; Fig. 2E). Because clicks are produced by an ultrafast and discrete strike of the drumming apparatus (Cook et al., 2024), the resulting sounds are broadband. When keeping the tank geometry, as well as the distance between the fish and hydrophone, stable, clicks show a similar frequency power spectrum across development (Fig. S2). Taken together, this suggests that the sounds of individual clicks are stable, but become louder with age.

We next investigated the development of the D. cerebrum sound production apparatus via bone and cartilage staining. In adults, the drumming apparatus and the skeletal structures associated with the anterior swim bladder are sexually dimorphic (Fig. 3A,B). Adult males have a hypertrophied 5th rib, a globular cartilage and two extra components of the os suspensorium (OS), a bone structure that supports the anterior swim bladder anteriorly and dorsally (Britz et al., 2021). The inner part of the male OS (iOS) has a posterior extension on the dorsal side that extends over the swim bladder. The male outer OS (oOS) extends further rostral and has an additional connection to the transverse process of the second vertebra. An additional feature of the sexual dimorphism is the position of the vent of the digestive tract, which is shifted anteriorly to the pelvic fin in males (Britz et al., 2021). This rerouting of the vent is the last sexually dimorphic developmental change in juvenile males, occurring after all components of the drumming apparatus have developed (Fig. 3A,B, bottom). However, the connecting flanges between the outer and inner arms of the OS, as well as the connection between the oOS and the vertebral column, continue to grow in adult males.

In 2- to 4-week-old juveniles, sexes cannot yet be distinguished morphologically. Undifferentiated juveniles resemble the adult female morphotype in their skeletal anatomy, with a thin 5th rib and only the lower extension of the iOS present (SL=7 mm; Fig. 3C). At 7.5 mm, the still undifferentiated juvenile develops the short, female version of the oOS (SL=7.5 mm; Fig. 3C). The first male-specific component that emerges at a SL of 8.0 mm is a hook-like outgrowth of the oOS, which starts to form around the same time of the increase in hypertrophy of the 5th rib (SL=8 mm; Fig. 3C). The next component is the posterior extension of the iOS over the swim bladder, which coincides with a faint first staining of the drumming cartilage (SL=8.8 mm; Fig. 3C). The final step in the emergence of the male drumming apparatus is the connection of the oOS to the vertebral column (SL=9.4 mm; Fig. 3C). The oOS continues to grow and becomes increasingly ossified during further maturation (Fig. 3A,B). In summary, the drumming apparatus components form in juvenile males at a SL between 8 and 10 mm. This is in line with the ontogeny of recorded clicks sounds in our behavioural experiments (Figs 1 and 2).

Danionella cerebrum is a miniature fish with a male-specific trait of acoustic signalling. Here, we showed that males start producing sounds after 1 to 1.5 months of development, at a SL of ca. 9 mm. In line with this, we found that the sound-producing drumming apparatus develops in male specimens between 7 and 10 mm of SL. A full set of drumming components is present at approximately 9 mm SL, while the OS continues to ossify and the rerouting of the vent is yet to occur. We also found that click repetition rate, sound profile and broadband frequency spectra were consistent across age, while sound intensity increased with maturation.

Acoustic allometry, a relationship between body size and sound features, has been described across species of primates and carnivore mammals (Bowling et al., 2017), as well as anurans (Gingras et al., 2013). Although within-species allometry is more complex in vocal learning species and possibly species with the capability of volitional sound modulation (Garcia and Ravignani, 2020), it has been described in several fish species with differing sound production mechanisms. For instance, larger fish of the cichlid species Metriaclima zebra produce lower frequency and louder sounds (Bertucci et al., 2012a), a trend that was also found when comparing juvenile and adult fish (Bertucci et al., 2012b). Similar results have been reported for Lusitanian toadfish, Halobatrachus didactylus (Vasconcelos and Ladich, 2008), grey gurnard, Eutrigla gurnardus (Amorim and Hawkins, 2005), mochokid catfish, Synodontis schoutedeni (Lechner et al., 2010) and croaking gourami, Trichopsis vittata (Wysocki and Ladich, 2001). A study performed on the weakfish Cynoscion regalis described a negative correlation between pulse duration and dominant frequency (Connaughton et al., 2000). By comparing individual sizes, the authors found that larger males produce louder sounds with a longer pulse duration and lower dominant frequency, without any clear change in pulse repetition rate. This is mostly in line with our results in D. cerebrum, where we found an increase in sound amplitude with size and age (Fig. 2A,D), yet no change in pulse repetition rate (Fig. 1F). Unlike in weakfish and other species, we found no evidence of changes in pulse duration or frequency spectra with maturation (Fig. 2C,E). Most sonic fish with a swim bladder mechanism, including the weakfish, use direct muscle contractions to compress the swim bladder and produce sounds. In contrast, D. cerebrum employ an indirect sound production mechanism where a drumming cartilage snaps out and strikes the anterior swim bladder upon muscle contraction (Cook et al., 2024). Therefore, unlike in weakfish, the strength or size of the sonic muscle in D. cerebrum seems less likely to influence pulse duration and thus the frequency spectrum. Here, we show a positive correlation between sound amplitude and SB height (Fig. 2D), while the temporal click profile remained similar (Fig. 2C,E). The correlation between SB height and click amplitude is unlikely to be a causal link, but rather a by-product of increased maturity of the structures of the drumming apparatus, such as the inner arm of the OS with its dorsal and ventral extension around the swim bladder and likely the mass of the sonic muscle. In line with this, a study comparing sonic muscle mass in the weakfish across spawning seasons showed a positive correlation with sound pressure levels (Connaughton et al., 1997).

Adult males produce clicks at repetition rates of ∼60 and ∼120 Hz by repeated unilateral or by alternating bilateral contractions, respectively (Cook et al., 2024). The presence of both repetition rates in juveniles suggests that the pattern generating neuronal circuitry that controls the alternating contractions is fully functional at the developmental onset of acoustic signalling. Similar results have been reported in weakfish, where pulse repetition rate did not differ with body size (Connaughton et al., 2000). Our dataset shows a trend for older fish to have proportionally more clicks at 120 Hz, as seen in the higher peaks of the histogram (Fig. 1E). This can be explained by the increasing length of 120 Hz bursts with age, of up to hundreds of clicks. In contrast, 60 Hz bursts typically do not comprise more than 10 clicks (Fig. 1J). Furthermore, we show that throughout all ages tested, 80% of all bursts consist of only either 60 or 120 Hz click repetition rates. However, younger ages show a higher likelihood of 60 Hz bursts over 120 Hz (Fig. 1I). We note that the high variability in burst types and durations between groups could be a result of not having the same sex ratio in each tank, owing to the repeated-measures design of the experiment. Given the positive relationship between maturation and sound amplitude, there is a possibility that younger fish produced sounds below the detection threshold of our setup. However, because our morphological analysis showed that the first cartilage staining appeared in fish with a SL of at least 8 mm (matching an age of 6–7 weeks in the repeated-measures sound recordings), it is unlikely that males are able to produce sounds earlier than that. Indeed, in the repeated-measures experiments first clicks appeared at 8.1±1.3 mm SL (Fig. 1B,D). Because there is individual variability in growth rate, the larger males in the group recordings could have been the ones first producing sounds. In the close-range click recordings with fewer fish per group and only males, body size measurements aligned well with the morphological results; the youngest clicks we recorded were at 9.3 mm SL (Fig. 2C), a stage at which all components of the drumming apparatus have formed (Fig. 3C).

To date, there is limited knowledge about the link between morphology of the sound producing machinery and the resulting sound features across ontogenetic development. Several studies have measured sound production throughout development in fish (Vasconcelos and Ladich, 2008; Amorim and Hawkins, 2005; Lechner et al., 2010), yet without morphological measures beyond body size. In contrast, a few studies have described the morphological development of sound-related structures (Hill et al., 1987; Fine, 1989; Brantley et al., 1993), but without measuring sound output. One exception is a study by Kéver and colleagues (Kéver et al., 2012) in Ophidion rochei. Unlike female D. cerebrum, females of O. rochei can produce sounds. The first juvenile sounds that were recorded in O. rochei resembled female sounds rather than adult male sounds (Kéver et al., 2012). Similar to D. cerebrum, the sound apparatus of O. rochei is sexually dimorphic and juvenile males initially resemble the female morphotype before differentiating new sound-producing structures during gonadal maturation (Kéver et al., 2012).

To date, our understanding of the neuronal processes underlying the ontogenetic development of sound production and acoustic communication is limited owing to methodological constraints of large-scale neuronal recordings across different developmental stages. Danionella cerebrum, with optical accessibility during its entire life, offers a unique opportunity as a vertebrate model to study the ontogeny of acoustic communication. An essential part of studying the development of acoustic communication is sound perception. Several studies have applied auditory brainstem recordings to probe hearing capabilities across age and related it back to the sound parameters individuals were able to produce (Lechner et al., 2010; Vasconcelos and Ladich, 2008; Wysocki and Ladich, 2001). The small, transparent D. cerebrum will enable whole-brain imaging of sound perception across development to establish not only when sounds are transduced to neural signals in the brainstem, but also when they start to trigger activity in brain regions that are associated with social behaviours across vertebrates, such as the preoptic area of the hypothalamus (Goodson, 2005). Notably, the Weberian apparatus shows an accelerated development in the Danionella species (Conway et al., 2021) and our morphological results also showed its fully ossified components, i.e. tripus and scaphium, prior to the development of any of the sexually dimorphic components of the sonic apparatus (Fig. 3C). This suggests that hearing develops prior to sound production in D. cerebrum and it will be an interesting future study to measure the perception of social sounds in non-sonic juveniles. Furthermore, an ancestrally shared development of the neuronal vocal-motor circuits between fish and vocal terrestrial mammals has been proposed (Bass et al., 2008). Therefore, D. cerebrum may serve as a suitable model to study both the development of sound production circuitry, as well as the auditory perception of acoustic social signals.

We thank Johannes Veith, Maximilian Hoffmann, Jörg Henninger and Verity Cook for assistance with methodological implementation and discussions of the ongoing work. We also thank our fish facility team, especially Anne Wrana and Nadine Kroworz, for fish husbandry and Fine Stuhr-Wulff for assistance with water changes. We thank Ralf Britz and Andrew Bass for feedback on the manuscript.